Chapter 1 — Definition, Landscape and Stakes

You have an idea. It came to you at work, in a line-up, or in a spreadsheet you built because no tool did what you needed. You have told three people, and two of them said it was great. What you do in the next ninety days will decide whether the idea becomes a business or a story you tell about the business you almost started.

What You'll Be Able to Do

• Define a business model rigorously, and separate it from an idea, an opportunity, a business plan and a strategy.
• Describe the landscape a new business enters in 2026: how many businesses start and survive in Canada, what it now costs to test an idea, and where the money and the help come from.
• Explain why most ideas fail, using the evidence rather than folklore, and name the kind of evidence that prevents each failure.
• State what you need from the work of this book: decision-grade evidence on desirability, feasibility and viability, and a model that can be funded and executed.
• Name your constraints — time, money, skills, legal rules, personal risk — so that the methods of the following chapters are sized to your situation.
• Read this book efficiently, knowing which chapter answers which question and how each chapter connects to La Machine.

Why It Matters

A business model is the difference between an idea and a business. The idea is what you would like to exist; the model is how it will create value for a specific customer, deliver that value and capture enough of it to keep going. Most ventures do not fail because the founder lacked talent or effort. They fail because a model that was never tested met a market that had not been asked. For the founder, the stakes are a year of life and the savings that go with it. For the team, the partner or the family who joins the venture, the stakes are the same, multiplied. The work of this book is cheap compared with either.

In This Chapter

• 1.1 What a Business Model Is — and Is Not
• 1.2 The Landscape in 2026
• 1.3 Why Ideas Fail
• 1.4 What the Founder Needs from This Work
• 1.5 The Founder's Constraints
• 1.6 How This Book Is Organized

1.1 What a Business Model Is — and Is Not

Ask ten founders what their business model is and you will hear ten different kinds of answer: "subscription", "we sell to restaurants", "it's like Uber for dog walkers", "we haven't decided yet", or a forty-page document. Each answer names a piece of a model and mistakes it for the whole. The confusion is expensive, because a founder who believes that "subscription" is a business model will spend months on billing software before discovering that nobody wants to subscribe.

1.1.1 The definition

DEFINITION Business model A business model "describes the rationale of how an organization creates, delivers, and captures value" (Osterwalder & Pigneur, 2010, p. 14). In plain terms: a business model is the set of choices that explain who your customers are, what they get from you that they value, how it reaches them, and how the money you collect exceeds the money you spend — repeatedly, and at a scale worth the effort. It is not a document but a logic; the documents (a canvas, a plan) are ways of writing that logic down so that it can be tested.

Three ideas in that definition deserve a closer look, because each one corrects a common error.

Creating value. The model begins with a customer and a problem, not with a product. Peter Drucker argued that every organization runs on a "theory of the business": a set of assumptions about its environment, its mission and the competencies it needs, and that the theory works only when those assumptions fit reality, fit each other and are tested constantly (Drucker, 1994). A founder who cannot say which customer will get which value from the venture does not have a business model yet; they have a product idea.

Delivering value. The model includes the path from the offer to the customer: the channels, the relationship, the resources and the partners it takes to deliver. Many ideas that would be valuable are not deliverable by the person who had them, at the cost they can afford, in the time they have. That is a model problem, not a product problem, and it is discovered by asking how delivery will actually happen.

Capturing value. The model ends in money that stays. Joan Magretta compared a business model to a story that answers the old questions every manager must answer: who is the customer, what does the customer value, how do we make money in this business, and what is the underlying economic logic that explains how we can deliver value to customers at an appropriate cost (Magretta, 2002). A story that ends with "and then we will figure out how to charge" is not finished.

1.1.2 The nine questions a model answers

The most widely used way to write a business model down is the Business Model Canvas, which Chapter 6 presents in full. Before the canvas, it helps to hear its nine blocks as nine questions, because a founder can answer them in a conversation long before drawing anything. Who are the customers, and which group comes first? What do they get from us that they value more than the alternatives? Through which channels do they learn about us, buy and receive the offer? What kind of relationship do they expect — self-service, personal assistance, a community? How does the money come in, from whom, and for what? What resources must we own or control — people, equipment, data, a brand, money? What activities must we perform well — making, selling, solving problems, running a platform? Who are the partners and suppliers without whom the model does not work? And what does all of that cost, in fixed costs and in costs that grow with each sale?

A founder who can answer all nine questions in a coherent way has a business model. A founder who can answer them with evidence rather than with hopes has a validated business model. The distance between the two is the subject of this book.

1.1.3 What a business model is not

The neighboring notions are real and useful, but they are not synonyms, and confusing them costs time. Table 1.1 separates five notions that founders often use interchangeably. For each one it states the question it answers, the form it usually takes, the horizon it works on, the kind of evidence that supports it, and the book of this library that develops it. Read it as a progression: an idea becomes an opportunity when the problem and the market are identified, an opportunity becomes a business model when the nine questions are answered, a model becomes a plan when it is turned into forecasts and actions, and a strategy is the set of choices that keeps the model defensible over years.

Table 1.1 — Idea, opportunity, business model, business plan and strategy: what each one is

Two of these boundaries matter most for a founder.

Model versus plan. A business plan is written to be read by someone who must decide — a lender, an investor, a partner, a board — and it assumes that the model underneath it is sound. Writing the plan before the model is validated produces a beautifully formatted set of guesses; the forecasts inherit every untested assumption and multiply it by twelve months. This book stops where the plan begins, and Book A2 assumes you have done the work described here.

Model versus strategy. A strategy describes how the venture will remain valuable when competitors respond, customers change and technology moves. It matters enormously — but not yet. A venture that has not found a customer does not need a five-year defense; it needs to know whether anyone will pay. Chapter 8 borrows enough strategic analysis to choose an entry position, and Book A3 takes over from there.

KEY CONCEPT A model is a set of hypotheses until evidence says otherwise Every box of a business model, at the start, is a guess: that these customers have this problem, that they will pay this price, that this channel reaches them, that this cost is what it will cost. The founder's job in the first months is not to defend the guesses but to find out which ones are true, cheaply, before money and time are committed on the false ones. Steve Blank called this customer development, and stated its premise bluntly: no business plan survives first contact with customers (Blank & Dorf, 2012). Eric Ries generalized it into a method — build the smallest thing that can test a hypothesis, measure, learn, and repeat (Ries, 2011). Everything in this book is a way of turning a guess into evidence.

DEFINITION Hypothesis A hypothesis is a statement about customers, the offer or the numbers that can be proved false by evidence you can collect within your means. "Small accounting firms need a better client portal" is a belief; "at least six of ten owners of accounting firms with five to twenty employees will say, unprompted, that sharing documents with clients costs them more than two hours a week" is a hypothesis, because it names who, what, how many and a threshold. Chapter 10 shows how to extract the hypotheses hidden in a model and rank them by risk; Chapters 3 and 9 show how to test them.

1.1.4 The three fits

A business model succeeds when three different things are true at once, and a founder can check them separately. Desirability asks whether customers want the offer enough to change what they do and pay for it. Feasibility asks whether you can build and deliver it — with your skills, your partners, your technology and your money. Viability asks whether the money that comes in exceeds the money that goes out, per customer and in total, soon enough to survive. These three lenses come from design practice and were placed at the center of business model testing by David Bland and Alexander Osterwalder, who add a fourth, adaptability, for the model's fit with the wider environment — regulation, competition, trends (Bland & Osterwalder, 2020).

The three fits are useful because failures cluster by fit. A desirable, viable offer that the founder cannot deliver is a feasibility failure. A feasible, desirable offer that loses money on every sale is a viability failure. And the most common failure of all — a feasible, viable offer for a problem nobody has — is a desirability failure. Figure 1.1 places the fits on the path this book follows, from the idea to the validated model and, beyond it, to the plan.

Figure 1.1 — From idea to validated model. An idea becomes a set of hypotheses about desirability, feasibility and viability; evidence from interviews, tests and first numbers confirms or refutes them; the surviving hypotheses form a validated model, which the business plan (Book A2) then turns into forecasts and actions. Refuted hypotheses loop back as a pivot. Original figure.

1.2 The Landscape in 2026

The definition of a business model has been stable for two decades. The landscape a new business enters — how many are created, how many survive, what it costs to test an idea, where the money and the help come from — changes every year. This section describes the terrain as of September 2026, so that the following chapters can concentrate on principles that outlast it.

1.2.1 Most businesses are small, and most new businesses are tiny

As of December 2024, Canada counted 1,099,521 employer businesses. Of these, 1,079,188 (98.2%) were small businesses with 1 to 99 employees, 16,953 (1.5%) were medium-sized with 100 to 499 employees, and 3,380 (0.3%) were large with 500 or more. More than three out of four Canadian businesses (77.3%) have 1 to 9 employees, and micro-enterprises of 1 to 4 employees alone represent 59.1% of the total. Small businesses employed 5.8 million people in 2024, or 46.6% of the private-sector workforce (Innovation, Science and Economic Development Canada, 2025).

For a founder, these numbers carry two messages. The first is that the business you are about to start will, in all likelihood, be one of the 59.1%: a micro-enterprise where the founder is also the salesperson, the bookkeeper and the customer-service department. The methods of this book are sized for that reality, and every chapter says what changes when there is a team. The second message is that almost everything written about business models — the books, the conference talks, the software demos — is written for the venture-backed technology start-up, which is a tiny fraction of the population. The principles transfer; the scale does not. A café, a consulting practice, a specialty manufacturer and a software product all need a validated model, but they will not test it the same way or at the same cost.

1.2.2 Entry, exit and survival

Business creation in Canada is a large, steady flow. Over the years 2017 to 2021, an average of 18,480 employer businesses were created each year in the goods-producing sector and 62,550 in the services-producing sector — about 81,000 new employer businesses a year — while 16,880 and 57,629 respectively exited, about 74,500 a year (Innovation, Science and Economic Development Canada, 2025). Behind those flows sits a survival curve that every founder should know before starting. Of the employer businesses created in Canada, 68.0% were still active five years after entry (69.7% in the goods-producing sector, 67.5% in services), and 48.2% were still active after ten years (52.0% in goods, 47.2% in services) (Innovation, Science and Economic Development Canada, 2025).

Read those numbers plainly: about one employer business in three is gone within five years, and about one in two within ten. Read them carefully, too. An exit is not always a failure; a business can be sold, merged or closed by an owner who retires, and the statistics do not distinguish. But the curve is steep enough in the first years to make the founder's question precise: what separates the two thirds that reach year five from the third that do not? Section 1.3 gives the answer the evidence supports, and it is not luck.

1.2.3 The recent drought

The years 2024 and 2025 were unusual. According to an analysis of Statistics Canada's business dynamics data by the Canadian Federation of Independent Business, more businesses exited than entered the market from early 2024 onward, with the gap widening in 2025: the exit rate stood at 5.6% in the second quarter of 2025 while the entry rate fell to 4.8% by the fourth quarter — among the weakest start-up activity and highest closure rates recorded outside the pandemic. The same analysis notes that business entry rates have fallen by nearly half since the mid-1980s (Canadian Federation of Independent Business, 2026). These are a federation's readings of public data, published in April 2026, and they should be checked against the current Statistics Canada release before being quoted; the direction, however, is consistent with the survival curve above. A founder starting in 2026 enters a market with fewer new competitors than in previous decades, and with customers who have seen suppliers close. Evidence of durability — a model that works on paper and in the first thirty customers — is worth more than it has been in a generation.

1.2.4 The cost of testing an idea has collapsed

Twenty years ago, finding out whether people wanted a product meant building it. Today a founder can describe an offer on a web page in an afternoon, put it in front of a few hundred people the same week, take pre-orders with a payment link, and run the service by hand for the first ten customers before writing a line of software. No-code tools, cloud services, online payments, scheduling pages and form builders — most of them free at the scale of a test — have turned validation from an expensive luxury into the cheapest phase of a venture. Chapter 9 catalogs the tests; Chapter 13 catalogs the tools.

The adoption of these tools by existing businesses remains uneven, which is itself an opportunity. Statistics Canada's Survey of Digital Technology and Internet Use found that in 2023, 48% of Canadian businesses with five or more employees used cloud computing, 32% received e-commerce orders, and only 7% used software or hardware with artificial intelligence — 6% of small businesses against 26% of large ones (Statistics Canada, 2024). A new venture that is systematic about testing and digital from the first day is not catching up with incumbents; in most local and business-to-business markets it starts ahead of them.

WHAT CHANGED The cost of a test As of 2026, the marginal cost of testing a business idea before building it is close to zero in tools and a few hours in time: a landing page with a sign-up form, a scheduling link for interviews, a payment link for pre-orders, a spreadsheet for the numbers and an assistant that drafts the first versions of all of them. The change began with no-code website and form builders around 2015, accelerated with online payments and scheduling around 2020, and became decisive in 2023–2025 when AI assistants made first drafts, summaries of interviews and simple analyses nearly free. The consequence is not that validation is easier to skip; it is that skipping it is now harder to excuse.

1.2.5 Where the money and the help come from

A founder in Canada rarely starts alone or unfunded, even when it feels that way. The financing ladder — personal savings, customers who pay in advance, loans, grants, equity — is the subject of Book A5; the point here is that most of it is open only to a founder who can show a model with evidence. Lenders and support programs read plans, and plans are built on models. The Québec and Canada Note below names the organizations a founder should know on day one, as of September 2026.

QUÉBEC & CANADA NOTE Starting a business in Canada: who to talk to first (as of September 2026) Québec.ca — Entreprises et travailleurs autonomes and Canada.ca — Business and industry are the official entry points for registrations, permits, taxes and programs; Book A4 walks through the legal steps. BDC (Business Development Bank of Canada) offers financing and advisory services and publishes free guides and templates for start-ups. Futurpreneur Canada finances founders aged 18 to 39 at the pre-start-up or start-up stage: a Futurpreneur loan of up to $25,000 CAD combined with up to $50,000 from BDC, for a total of up to $75,000, with up to two years of one-to-one mentorship; the Futurpreneur portion carries interest at the CIBC prime rate plus 3% (capped at 9%) with interest-only payments in the first year, and the applicant must show training or experience in the field (Futurpreneur Canada, 2026). Investissement Québec finances and supports Québec businesses; PME MTL and the network of local development centers, and the SADC and CAE in the regions, offer advisory services and small loans; incubators and accelerators (Chapter 12) add programs. Equity crowdfunding is regulated: under National Instrument 45-110, a start-up may raise up to $1,500,000 in a twelve-month period through a registered funding portal, and an investor may contribute up to $2,500 per distribution, or $10,000 if advised by a registered dealer (Canadian Securities Administrators, 2021). Every one of these doors opens more easily to a founder who arrives with evidence rather than with enthusiasm.

1.3 Why Ideas Fail

The landscape explains what happens on average. The evidence on failure explains why, and it points to a small number of causes that the work of this book is designed to address.

1.3.1 What the post-mortems say

The most cited body of evidence on start-up failure is the collection of post-mortems assembled by the research firm CB Insights. Its March 2026 analysis reviewed 431 venture-backed companies that shut down since 2023 and, for the 385 whose reasons could be identified, found that 70% cited running out of capital, 43% cited poor product-market fit, 29% cited bad timing or macroeconomic conditions, and 19% cited unsustainable unit economics; most companies cited more than one reason, so the percentages exceed one hundred (CB Insights, 2026). In earlier editions of the same analysis, the leading cause was stated as "no market need", cited in roughly two fifths of cases — a different label for the same finding.

Two cautions apply. These companies had raised venture capital, which most small businesses never do; their failures are better documented than those of the café on the corner, not necessarily different. And "ran out of capital" is a symptom, as the analysis itself notes: money runs out because customers did not come, or did not stay, or did not pay enough. Read through the symptom, the causes line up with the three fits of Section 1.1.4. Poor product-market fit is a desirability failure. Unsustainable unit economics is a viability failure. Bad timing is a failure of adaptability — the environment was not ready, or moved. Running out of capital is what all of them look like from the bank account.

1.3.2 The solution in search of a problem

The single most common path to a desirability failure is so well worn that it has a name: the solution in search of a problem. A founder has a skill, a technology or an inspiration; builds something with it; and then looks for the people who might want it. The order is backwards, and it fails for a structural reason. When the solution comes first, every conversation with a potential customer becomes a sales conversation, in which the founder listens for confirmation and the customer, who is polite, provides it. The founder hears "that's interesting" as "I would pay for that." Twelve months and a product later, the founder discovers that interest and payment are different things.

The remedy is not to have fewer ideas but to hold them differently. Chapter 2 turns the order around: it starts with problems, scores them for frequency, intensity and willingness to pay, and only then asks which solution fits. Chapter 3 gives the conversation a structure — the customer's past behavior, not their opinion of your future product — that makes politeness harmless.

CASE IN POINT A software idea that skipped discovery Fictional example; names and figures are invented. Two friends, a developer and a logistics coordinator, noticed that parents in their suburb organized car-pooling for children's sports in group chats that constantly broke down. They designed Rideline, an application that matched families by schedule and route, and spent fourteen months and about $62,000 CAD — savings, evenings, a contractor for the mobile version — building it before showing it to anyone outside their circle. The launch reached 1,100 downloads through local parent groups; 140 families created a profile; 9 ever completed a ride through the app. Interviews conducted afterwards, when it was too late, were unanimous: parents did not want a better way to organize car-pooling with strangers, because the whole point of the group chat was that they knew and trusted the other families. The problem was real, frequent and intense, but the job was trust, not scheduling — and no amount of scheduling software does that job. Twelve interviews in the first month would have found it. The founders did not lack talent; they lacked the order of operations that this book imposes.

1.3.3 The other causes

Desirability is the first cause but not the only one. Viability failures are quieter: the offer sells, and every sale loses money once the real cost of acquiring and serving a customer is counted; Chapter 7 gives the arithmetic that exposes them on paper. Feasibility failures appear when the founder cannot deliver what was sold — the skill is missing, the supplier does not exist at that price, the technology takes three times longer; Chapter 6 forces the left side of the canvas to be as explicit as the right, and Chapter 12 names the help. Timing failures are the hardest to prevent and the easiest to detect early: a test that shows no urgency in the market is a timing signal, and Chapter 2 treats "why now" as a question with an answer. Competitive failures — an incumbent responds, a substitute is good enough — are addressed in Chapter 8, whose first lesson is that the founder's most common competitor is the customer doing nothing at all. And team failures, which Noam Wasserman's study of thousands of founders shows to be among the most frequent causes of trouble, are the subject of Chapter 12 (Wasserman, 2012).

1.3.4 The cost of skipping validation

The reason to validate is not virtue; it is arithmetic. The Do the Math box compares, with illustrative figures, a year spent building on untested assumptions with ninety days spent testing them.

DO THE MATH The cost of a year without validation Illustrative figures. A founder keeps a day job and works on the venture fifteen hours a week for a year. Valued at a modest $75 CAD an hour, that time is worth 15 × 50 × $75 = $56,250. Add a contractor for the parts the founder cannot build ($18,000), tools, incorporation and legal ($6,000), and a launch campaign ($4,000): the year costs about $84,000 in money and displaced income before the first real customer conversation. Compare the discovery work of this book: six hours a week for thirteen weeks (78 hours, $5,850), a designer for a landing page and a prototype (two days, $1,400), tools at the scale of a test ($300) and a lawyer's hour on pre-order terms ($350): about $7,900, or less than a tenth. The discovery quarter does not guarantee success; it guarantees that the $84,000 year, if it happens, is spent on a model whose riskiest assumptions have already survived contact with customers. And when the idea is a Rideline, the quarter saves the year entirely.

1.3.5 What failure is not

An invalidated hypothesis is not a failure; it is the product the discovery process is designed to produce. The founder who learns in week three that restaurant owners will not pay for a scheduling tool, and in week six that they will pay for a no-show reduction service, has not failed twice — they have found a model. Chapter 10 gives the vocabulary (validated, invalidated, inconclusive; persevere, pivot, stop) and the discipline to make those decisions on evidence. The failure this book is written to prevent is the other kind: the year spent, the money gone and the lesson learned only at the end.

1.4 What the Founder Needs from This Work

A founder who reads this book is looking for something specific, even if it is not yet named. This section names it, in five deliverables that the following chapters produce and that Chapter 14 sequences over ninety days.

1.4.1 Five deliverables

A problem worth solving, with evidence. Not "people hate doing their bookkeeping" but a written statement of who has the problem, how often, how intensely, what they do about it today and what it costs them — supported by interviews you conducted and data you can point to (Chapters 2 and 3).

A customer segment you can reach. A first group of customers, described precisely enough that you could list fifty of them by name or by source, sized honestly, and chosen because it is the easiest place to win rather than the biggest (Chapter 5).

A value proposition that has been tested. A statement of what the customer gets, why it is better than their alternative and what proof exists — tested with the people it describes, in their words (Chapter 4).

A model whose first numbers work. The nine blocks filled, the revenue mechanism chosen, the unit economics computed with first estimates and stress-tested: a contribution margin, a cost to acquire a customer, a payback period, a break-even, a runway (Chapters 6 to 8).

A decision with a log. After the tests of the riskiest assumptions (Chapters 9 and 10), a documented decision to persevere, pivot or stop — with the evidence, the reasoning and the date — that you, a partner or a lender can read later and understand.

1.4.2 Decision-grade evidence

The word evidence recurs in every chapter, so it should be defined once. Decision-grade evidence is evidence strong enough to justify committing money and time that you cannot get back. It has three properties. It is behavioral where possible — what people did, paid or signed, rather than what they said they might do. It is specific — a count, a rate, a quotation with a name and a date, not an impression. And it is sufficient — enough observations that a pattern is a pattern and not two anecdotes. Chapter 3 draws the evidence ladder, from "says" to "does" to "pays", and Chapter 10 shows how much evidence a decision needs, depending on what the decision commits.

WHO DOES THIS Owning discovery In a solo venture, the founder owns every step of discovery and cannot delegate the customer conversations; what they can delegate is production (a landing page, a prototype, a survey layout) and what they can automate is logistics (scheduling, sign-ups, reminders). The minimum is one fixed block of six hours a week, protected from the day job. In a founding team, one person owns the customer conversations and the validation log while the others own feasibility (can we build it?) and viability (does the money work?), and the team meets weekly on the evidence rather than on opinions; the split of decisions, equity and roles is a Chapter 12 matter and is best settled early. In an established business launching a new offer or a new line, an intrapreneur owns discovery with a small budget and an explicit mandate, reports to the owner or the general manager on evidence at fixed gates, and is protected from the pressure to "just launch it" that an existing sales force applies. In all three cases, the founder never delegates the choice of which problem to solve and for whom — that choice is the business.

IN LA MACHINE Decisions that outlive the meeting La Machine keeps a registry of decisions at Arbitrages (/arbitrages): a question, the options, the recommendation, the decision and its date, kept forever and never closed without a written answer. A venture in discovery produces decisions every week — which segment first, which price to test, whether to persevere after a failed test — and the registry is where they belong, so that the reasoning survives the founder's memory and can be reread when the next decision arrives (Chapter 10, Section 10.7).

1.4.3 What investors, lenders and partners read

A founder rarely starts alone. A lender reads the plan for the ability to repay; an investor reads it for the size of the opportunity and the team's capacity to learn; a co-founder or a first employee reads the model for whether it is worth their next two years. All of them are reading for the same underlying thing: evidence that the model has met customers and survived. The deliverables of Section 1.4.1 are the raw material of Book A2's business plan and Book A5's financing file; produced in this order, they make those books short.

1.5 The Founder's Constraints

The landscape explains what is possible. The founder's situation explains what is hard. Six constraints come up in almost every new venture, and naming them at the start is the difference between a plan that fits the founder's life and a plan abandoned in week five.

1.5.1 Time

Most founders start with a job, a family or both. The realistic budget is six to ten hours a week — not forty — and it is the scarcest resource of the venture. The methods of this book are sized to six hours a week for a solo founder, and Chapter 11 lays them out in a two-week sprint that fits that budget. The constraint has one consequence that founders resist: with six hours a week, you cannot both build and validate. Validate first; the building goes faster afterwards, because it builds the right thing.

1.5.2 Money

The money available at the start is usually personal: savings, a spouse's income, a line of credit, family. Two facts about it shape everything. The first is that it is finite and that the founder's exposure is real — a personal guarantee on a loan or a lease follows the person, not the business, and Book A4 explains how the legal form limits some of that exposure and not all of it. The second is that money spent before validation buys the wrong things at full price: a brand, a website, an inventory, a build, for a model that may change three times. Chapter 7 introduces the runway — the number of months the venture can survive at its current burn — as the founder's clock, and Chapter 14's plan assumes a discovery budget of a few hundred to a few thousand dollars, not tens of thousands.

1.5.3 Skills

Chapter 12 describes six skill families — customer, product, sales, finance, operations, legal — and offers a self-assessment grid. No founder has all six, and the honest grid is the first step toward the team, the advisors and the professionals who fill the gaps. The most common gap is not technical; it is the customer skill of asking questions and hearing answers you do not like. It can be learned in a month with the script of Chapter 3.

1.5.4 Legal and regulatory

A founder in discovery is not yet running a business, but several rules apply from the first customer conversation, and they are design inputs rather than obstacles. Personal information collected during research — names, emails, interview notes — is protected: in Québec, the Act respecting the protection of personal information in the private sector (as reformed by Law 25) requires that a person be informed of the purposes of collection and consent to them, and federally the Personal Information Protection and Electronic Documents Act applies similar principles; Chapter 3 gives the practical minimum. Electronic messages to a waiting list are commercial electronic messages under Canada's anti-spam legislation, which requires consent, identification and an unsubscribe mechanism (CASL s. 6); Chapter 9 shows how to collect sign-ups lawfully. Claims made on a test page or in a pre-sale are subject to the Competition Act's prohibition of false or misleading representations (s. 74.01), whether or not the product exists yet. Pre-orders from consumers are distance contracts under Québec's Consumer Protection Act, with disclosure and cancellation rules that Chapter 3 summarizes. Taxes begin with the first sale: a business must register for the GST and the QST once its worldwide taxable supplies exceed $30,000 in a calendar quarter or over the four preceding quarters, and may register voluntarily before (Revenu Québec, 2026); registration with the Registraire des entreprises, permits and licenses are the subject of Book A4. None of these rules prevents a test; each one shapes how the test is run.

1.5.5 Personal risk and the people around you

A venture consumes evenings, savings and attention, and the people who share them are stakeholders whether or not they are on the canvas. Wasserman's research on founders found that the choices made early — going solo or with co-founders, splitting equity, deciding who controls what — predict much of the later conflict (Wasserman, 2012). The discovery quarter is the cheapest moment to have those conversations, because nothing has been built and nobody has yet sacrificed a year. Chapter 12 gives the questions; Chapter 10 treats personal risk alongside the model's risks in the risk register.

1.5.6 Technology and tools

The tools of discovery are mostly free or already paid for, and Chapter 13 lists them by function. The constraint is not access but discipline: a founder can spend the discovery quarter comparing website builders instead of talking to customers. The rule of this book is that a tool is adopted when a task requires it and not before, and that the five tools of day one — a note-taking place for the validation log, a form, a scheduling page, a spreadsheet and a contact base — cover the whole of the first month.

WATCH OUT Spending before evidence The purchases a founder makes in the first weeks — a domain and a brand identity, a website, a trademark search, inventory, a development contract, a lease — feel like progress and are recorded as such. Each of them commits money to a version of the model that discovery will change. The typical cost of the premature list above is $8,000 to $25,000 CAD in a small venture, and its worst effect is not the money: it is that the founder, having spent it, stops being able to hear that the model is wrong. Buy the domain if you must; delay everything else until the riskiest assumptions of Chapter 10 have been tested.

1.6 How This Book Is Organized

1.6.1 The spine

Every book in this library follows the same spine, so that a reader always knows where to look. This chapter has defined the business model, described the landscape, explained why ideas fail and named the founder's needs and constraints. Chapters 2 to 10 teach the concepts, rules and methods of the subject, in the order in which a founder uses them: spotting and shaping an opportunity (2), discovering and validating with customers (3), writing the value proposition (4), sizing the market and choosing the first segment (5), drawing the business model on the canvas (6), doing the first numbers (7), analyzing competitors and choosing an entry position (8), testing with prototypes, minimum viable products and pilots (9), and managing assumptions, risks and pivots (10). Chapter 11 turns the method into tasks with a cadence and verification criteria; Chapter 12 describes the skills, roles and help; Chapter 13 lists the tools by function; Chapter 14 gives the advice, the pitfalls, the ten numbers to watch and a ninety-day plan.

1.6.2 The boxes

Twelve kinds of boxes punctuate the text, each labeled so that you can find what you need: a Key Concept to retain from a section; a Definition of a term, collected in the glossary; The Rule for an obligation, with its source and its jurisdiction; a Method with steps, a deliverable and its pitfalls; Do the Math for a fully worked calculation with illustrative figures; a Case in Point set in a fictional small business; Watch Out for a costly pitfall; Tools of the Trade for what a tool must do and popular examples, dated; Who Does This for roles in the three situations; In La Machine for the page or workflow that supports the point; What Changed for a recent change with its date; and the Québec & Canada Note for a local rule, threshold or organization, dated and sourced.

1.6.3 The closing sections

Every chapter ends the same way: the Costly Mistakes of the subject with their consequences, the Key Takeaways, Your Action Plan with deliverables, timings and verification criteria, In La Machine — how the software supports the chapter and honestly what it does not do yet — a Self-Check with its answers, and the Chapter References in APA style. If you have twenty minutes for a chapter, read the opener and the closing sections; the body is there when you need the mechanism.

1.6.4 The three situations

Throughout, methods are checked against three situations. The solo founder does everything, so simplicity and automation dominate. The founding team divides the work, so definitions, logs and weekly reviews dominate. The established business launching something new has resources and a sales force but also inertia and internal customers to satisfy, so mandates, gates and protection from premature scaling dominate. When the advice differs by situation, the difference is stated — often in a table with those three columns.

1.6.5 Where this book stops

This book explains how to go from an idea to a validated business model. It deliberately stops before the documents and decisions that follow: the business plan and its forecasts (Book A2), the strategy that keeps the model defensible (A3), the legal form, registration and permits (A4), the financing of the venture (A5), the detailed design of the offer (A6) and the pricing strategy (A7). It borrows from marketing — the ideal customer profile and the positioning statement were defined in Book B1 and are used here for a venture without customers yet — and it does not operate any channel: marketing fundamentals live in B1, market research in B2, the sales process in E1, and building the product or the website in K1 and C2. Where a chapter touches those subjects, a cross-reference (→ see Book …) points the way.

Costly Mistakes

1. Confusing the idea with the model. The founder who says "our business model is subscription" has named a revenue mechanism and skipped the eight other questions. The consequence is months spent on billing, branding and a build for an offer whose customer, value and channel were never defined — typically $20,000 to $80,000 CAD and a year, as in the Do the Math box above.
2. Writing the plan before validating the model. A forty-page business plan built on untested assumptions inherits every one of them and multiplies it by twelve months of forecasts. The consequence is a financing file that a lender or an investor reads as a set of guesses, and a founder who defends the guesses instead of testing them.
3. Building the solution first and looking for the problem afterwards. Every customer conversation becomes a sales conversation, politeness is read as demand, and the product meets the market only at launch. The consequence, in the post-mortem evidence, is the leading cause of failure: no market need, discovered too late.
4. Reading the survival curve as luck. Treating the one-in-three exits before year five as fate rather than as the predictable result of untested models leads founders to skip the cheapest phase of the venture. The consequence is entering the statistic on the wrong side.
5. Spending before evidence. Brand, website, inventory, development contract and lease bought in the first weeks commit $8,000 to $25,000 CAD to a version of the model that discovery will change — and, worse, make the founder unable to hear that it must change.
6. Ignoring the rules that apply from the first conversation. Collecting emails without consent, promising on a test page what cannot be delivered, taking consumer pre-orders without the required disclosures: each one is cheap to do correctly and expensive to undo, and each one is explained in Chapters 3 and 9.

Key Takeaways

• A business model is the logic by which a venture creates, delivers and captures value. It answers nine questions about customers, value, channels, relationships, revenue, resources, activities, partners and costs; a founder who can answer them coherently has a model, and one who can answer them with evidence has a validated model.
• An idea, an opportunity, a model, a plan and a strategy are different things. They form a progression, and doing them out of order — the plan before the model, the strategy before the customer — produces documents rather than businesses.
• Every model starts as a set of hypotheses. The founder's first job is to find out which ones are true, cheaply, before committing money and time on the false ones; that is what customer development and the lean start-up method are for.
• Three fits must hold at once: desirability, feasibility and viability, with adaptability as the fourth. Failures cluster by fit, and the most common failure is a feasible, viable offer for a problem nobody has.
• The landscape favors the systematic founder. About one employer business in three exits within five years and one in two within ten; the cost of testing an idea before building it has collapsed; and lenders, programs and partners open more easily to evidence than to enthusiasm.
• Failure has causes that can be tested for. Running out of capital is the symptom; no market need, unsustainable unit economics, bad timing, an unfeasible delivery and a fractured team are the causes, and each one has a chapter.
• Constraints are design inputs. Six hours a week, a few thousand dollars, an honest skills grid, the rules on personal information, electronic messages, claims, pre-orders and taxes, and the people around the founder all shape the plan — and a plan that ignores them is abandoned in week five.
• Never delegate the choice of which problem to solve and for whom. Everything else — production, logistics, analysis — can be delegated to a team member, a professional or a machine.

Your Action Plan

The plan below lists the five steps that close this chapter, in order. For each step it names what to do, the deliverable it produces, how long it should take and how you will know it is done. The first three steps produce the documents that every later chapter builds on; the last two set the conditions — the owner, the time, the decision registry — under which the work of the next ninety days can happen.

In La Machine

La Machine is built around tasks with verification criteria, projects that group them, and decisions that are recorded rather than remembered. For this chapter, four pages matter, and they are the same pages a founder will use throughout the discovery quarter.

• Today (/aujourdhui) is the daily screen of the plan, open to the administrator and agent roles. Its four tiles and seven cards show the runs waiting for approval, the overdue tasks, the count of tasks due today, the runs completed today and the recent failures, and it is where the weekly discovery slot begins.
• Projects (/projets) group the work of a venture, one card per project with a gauge of its completed and recurring tasks. No page creates a project, so the project for each idea under investigation is loaded by whoever administers the platform, and the tasks of the following chapters are added to it as they arise. The Projects page does not display the documents stored with a project, so throughout this book a document "attached to the project" is a link in the Documents section of one of the project's tasks, or a note attached to the project on the Notes page (/presse-papier); the nine-answer sheet, the hypothesis list and the constraints sheet of the action plan are attached that way to its first task.
• Tasks (/taches) carry the discovery work, each with a verification criterion that the platform requires at creation. The platform refuses a task whose criterion is missing, but it does not judge the wording, so the difference between "talk to customers" and "conduct six interviews with owners of accounting firms by the 20th, notes filed in the log" is the founder's to make: only the second can be checked by someone else. Chapter 11 describes the format in full.
• Arbitrages (/arbitrages) is the decision registry. Every choice of the discovery quarter — the first problem, the first segment, the price to test, whether to persevere after a failed test — is recorded there with its question, options, recommendation, decision and date, and no entry is closed without a written answer.

What La Machine does not do: it does not offer an interactive business model canvas or a value proposition canvas — the canvas of Chapter 6 is held as a project with notes and tasks, or on a whiteboard tool from Chapter 13. It does not run the customer interviews, choose the problem or decide which segment comes first; those are the founder's decisions, which the platform then tracks. Book N1 (Getting Started with La Machine) walks through the navigation, and Book N2 (Piloting Tasks, Executions and Projects) covers tasks, runs and projects page by page.

Self-Check

1. In one sentence, what is a business model, and what are the three things it must do?
2. Name the nine questions a business model answers.
3. What distinguishes a hypothesis from a belief, and why does the distinction matter for a founder?
4. According to the evidence cited in this chapter, what share of employer businesses created in Canada were still active after five years, and after ten (as of the 2025 statistics)?
5. Why is "ran out of capital" described as a symptom rather than a cause, and which three fits do the underlying causes correspond to?

Answers. (1) A business model is the logic by which a venture creates value for a specific customer, delivers it and captures enough of it to keep going (Section 1.1.1). (2) Who the customers are; what they value; through which channels; what relationship; how the money comes in; what resources; what activities; which partners; and what it all costs (Section 1.1.2). (3) A hypothesis names who, what, how many and a threshold, so that evidence within the founder's means can prove it false; a belief cannot be tested. The distinction matters because the discovery quarter exists to turn guesses into evidence before money is committed (Section 1.1.1). (4) 68.0% after five years and 48.2% after ten, according to the Key Small Business Statistics 2025 (Section 1.2.2). (5) Money runs out because customers did not come, did not stay or did not pay enough; the underlying causes — no market need, unsustainable unit economics, an undeliverable offer — are failures of desirability, viability and feasibility respectively, with bad timing as a failure of adaptability (Section 1.3.1).

Chapter References

An Act to promote the efficiency and adaptability of the Canadian economy by regulating certain activities that discourage reliance on electronic means of carrying out commercial activities, S.C. 2010, c. 23 (Canada's Anti-Spam Legislation), s. 6. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/E-1.6/

Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25). LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1

Competition Act, R.S.C. 1985, c. C-34, s. 74.01. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-34/

Consumer Protection Act, CQLR c P-40.1, ss. 54.1–54.16. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-40.1

Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/P-8.6/

Bland, D. J., & Osterwalder, A. (2020). Testing business ideas: A field guide for rapid experimentation. Wiley.

Blank, S., & Dorf, B. (2012). The startup owner's manual: The step-by-step guide for building a great company. K&S Ranch.

Canadian Federation of Independent Business. (2026, April 15). Canada's entrepreneurial drought — Part 1: The shrinking business landscape. Retrieved September 7, 2026, from https://www.cfib-fcei.ca/en/research-economic-analysis/canada-entrepreneurial-drought-part1

Canadian Securities Administrators. (2021). National Instrument 45-110 Start-up Crowdfunding Registration and Prospectus Exemptions. Ontario Securities Commission. Retrieved September 7, 2026, from https://www.osc.ca/en/securities-law/instruments-rules-policies/4/45-110

CB Insights. (2026, March 5). The top 9 reasons startups fail. Retrieved September 7, 2026, from https://www.cbinsights.com/research/report/startup-failure-reasons-top/

Drucker, P. F. (1994). The theory of the business. Harvard Business Review, 72(5), 95–104.

Futurpreneur Canada. (2026). Core startup program. Retrieved September 7, 2026, from https://futurpreneur.ca/en/offering/core-startup/

Innovation, Science and Economic Development Canada. (2025). Key small business statistics 2025. Government of Canada. Retrieved September 7, 2026, from https://ised-isde.canada.ca/site/sme-research-statistics/en/key-small-business-statistics/key-small-business-statistics-2025

Magretta, J. (2002). Why business models matter. Harvard Business Review, 80(5), 86–92.

Osterwalder, A., & Pigneur, Y. (2010). Business model generation: A handbook for visionaries, game changers, and challengers. Wiley.

Revenu Québec. (2026). Registering for the GST and QST. Gouvernement du Québec. Retrieved September 7, 2026, from https://www.revenuquebec.ca/en/businesses/consumption-taxes/gsthst-and-qst/registering-for-the-gst-and-qst/

Ries, E. (2011). The lean startup: How today's entrepreneurs use continuous innovation to create radically successful businesses. Crown Business.

Statistics Canada. (2024, September 17). Survey of Digital Technology and Internet Use, 2023. The Daily. Retrieved September 7, 2026, from https://www150.statcan.gc.ca/n1/daily-quotidien/240917/dq240917c-eng.htm

Wasserman, N. (2012). The founder's dilemmas: Anticipating and avoiding the pitfalls that can sink a startup. Princeton University Press.

Chapter 2 — Spotting and Shaping Opportunities

A bookkeeper notices that every one of her forty clients sends her the same shoebox of receipts in the same disorder every March, and that each of them has, at some point, tried and abandoned an app. A warehouse supervisor builds a spreadsheet to schedule forklift maintenance because the software the company bought cannot do it. Neither of them thinks of themselves as having a business idea. Both of them are standing on one.

What You'll Be Able to Do

• Recognize where opportunities come from — problems you have lived, skills you hold, access you own, trends you have noticed — and start from what you have rather than from what you lack.
• Score a problem for its quality on frequency, intensity, the cost of the current workaround, who pays, reachability and your own access, and tell a painkiller from a vitamin.
• Judge timing, by answering the question "why now?" with specific changes in technology, rules, demographics, behavior or economics.
• Use jobs-to-be-done as a lens to see the progress a customer is trying to make, and to find the jobs that existing offers serve badly or not at all.
• Spot market gaps in non-consumption, over-service, workarounds and complaints, and read the evidence each one leaves behind.
• Shape an idea into a one-page opportunity brief, and screen twenty ideas down to the one worth a discovery sprint.

Why It Matters

The idea a founder starts with is rarely the business they end up running, but the problem they start with usually is. A venture built on a problem that is frequent, intense and expensive for a reachable group of people can survive several wrong solutions; a venture built on a solution in search of a problem cannot survive one. For the founder, the work of this chapter decides where the next ninety days of effort go. For a team or an established business, it decides which of the many ideas on the table deserves the scarce hours of discovery — and which ones deserve a written reason for being set aside.

In This Chapter

• 2.1 Where Opportunities Come From
• 2.2 Problems Worth Solving
• 2.3 Trends and Timing
• 2.4 Jobs-to-Be-Done as an Opportunity Lens
• 2.5 Market Gaps and Underserved Segments
• 2.6 Shaping the Opportunity
• 2.7 Screening Many Ideas

2.1 Where Opportunities Come From

The popular image of the entrepreneur is a person struck by a bolt of inspiration. The evidence describes something more ordinary and more useful: most successful businesses start from something the founder already knew, did or had access to. Amar Bhidé's study of the founders of fast-growing American companies found that about seven in ten had replicated or modified an idea they had encountered in their previous employment, and that most had started with modest capital and adapted quickly rather than executing a grand plan (Bhidé, 2000). Scott Shane's review of the research on entrepreneurship reaches a related conclusion: opportunities are not evenly distributed, and the people who find them are the people whose prior knowledge, work and relationships put them in a position to see what others cannot (Shane, 2003).

2.1.1 Four sources

Opportunities come from four places, and a founder should take inventory of all four before looking anywhere else.

Problems lived. The bookkeeper's shoebox and the supervisor's spreadsheet are problems the founder has experienced personally, in a role they understand. These are the richest source, because the founder already knows the circumstance, the workaround, the cost and the people, and can find twenty more of them by tomorrow.

Skills held. A skill — a trade, a profession, a craft, a technical ability — is an opportunity when a group of people would pay to have it applied to their problem. The risk of starting from a skill is the solution-first trap of Chapter 1; the remedy is to list the problems the skill solves and score them (Section 2.2) before deciding what to sell.

Access owned. A founder may have access that others do not: a supplier relationship, a professional network, a licence, a language, a community, a body of data, a location. Access is an unfair advantage (Chapter 8) when it makes a problem cheaper to reach or a solution cheaper to deliver. It is one of the most underrated sources of opportunity, because it does not feel like an idea.

Trends noticed. A change in technology, regulation, demographics, behaviour or economics creates new problems and makes old solutions obsolete. Trends are the most exciting source and the most dangerous, because a trend is visible to everyone and the founder who has only noticed it has no advantage over the thousand others who noticed it too. Section 2.3 explains how to turn a trend into a timing argument rather than a business idea.

2.1.2 Start from what you have

Saras Sarasvathy's research on how expert entrepreneurs actually reason found that they rarely start from a goal and work backward to the means. They start from what they have — who they are, what they know and whom they know — and ask what they can do with it; they commit only what they can afford to lose; they treat surprises as material rather than as obstacles; and they build the venture with the people who commit to it along the way (Sarasvathy, 2001). She called this effectuation, in contrast with the causal reasoning of the business plan. For a founder in the first weeks, the practical consequence is a discipline: before evaluating an idea for how big it could become, evaluate it for how much of it you can start with what you already have, and how much you can afford to lose finding out whether it works.

DEFINITION Opportunity An opportunity is a problem that a specific, reachable group of people has often and intensely enough to pay for a solution, at a time when a solution has become possible or newly valuable, and in a position from which you can deliver it. An idea is a candidate for an opportunity; it becomes one when the problem, the people, the willingness to pay, the timing and the founder's access have each been checked against evidence. An opportunity is not a solution: several solutions can fit the same opportunity, and the first one imagined is rarely the best.

KEY CONCEPT Fall in love with the problem A founder who loves the solution defends it against every piece of evidence. A founder who loves the problem uses every piece of evidence to find the solution that fits. The difference is not temperament; it is the order of operations. Write the problem first — in the words of the people who have it, with its frequency, its cost and its current workaround — and treat every solution, including your first one, as a hypothesis to be tested against that problem. The chapters that follow assume you have done so.

2.2 Problems Worth Solving

Not every real problem is worth a business. People have thousands of problems they will never pay to solve, either because the problem is rare, or mild, or because the workaround is good enough, or because the person who feels the pain is not the person who holds the budget. This section gives the criteria that separate problems worth solving from problems merely worth noticing.

2.2.1 Frequency, intensity and willingness to pay

Three questions do most of the work. How often does the problem occur for one person — daily, weekly, monthly, once a year, once in a lifetime? How intense is it when it occurs — a mild irritation, a lost hour, a lost customer, a fine, a sleepless night? And would they pay — do the people who have the problem already spend money or time on a workaround, and is the person who feels the pain the same person who decides and pays?

The three multiply. A daily, mild, unpaid-for problem (a slow coffee machine) and a rare, intense, paid-for one (a flooded basement) are both weaker than a weekly, intense, budgeted one (a payroll that must be right every second Thursday). The strongest problems are frequent enough to build a habit, intense enough to justify a change, and already attached to money — a budget line, a supplier, a fee, a salary spent on the workaround.

2.2.2 Painkillers and vitamins

Practitioners distinguish painkillers — offers that remove a problem the customer already feels and already tries to solve — from vitamins, offers that promise an improvement the customer does not yet feel the lack of. The distinction is useful and it is not moral: vitamins can be excellent businesses. But they are harder businesses for a new venture, because they require the founder to create the sense of a problem before selling its solution, and creating demand is expensive. A first venture should almost always be a painkiller; a vitamin is what an established business adds once it has customers who trust it (→ see Book A6, Designing a Value-Added Offer).

2.2.3 The problem scorecard

Table 2.1 turns the criteria of this section into a scorecard that a founder can apply to any candidate problem in thirty minutes. It lists six criteria, each scored from 1 to 5, and for each one it states what a score of 1 and a score of 5 look like, so that two people scoring the same problem arrive near the same number. The six criteria are the frequency of the problem, its intensity, the cost of the current workaround, whether the person who feels the pain also pays, whether the people who have the problem can be reached at a bearable cost, and whether the founder has access that others lack. A total near 30 signals a problem worth the discovery sprint of Chapter 3; a total under 15 signals a problem to record and set aside.

Table 2.1 — The problem scorecard (six criteria, each scored from 1 to 5)

METHOD The problem scorecard in thirty minutes When to use it: on every candidate problem, before any interview, and again after the first six interviews. Steps: (1) Write the problem in one sentence in the words of someone who has it, not in yours. (2) Name one person you know who has it; if you cannot, score reachability and access at 1 and note it. (3) Score the six criteria of Table 2.1 from what you know today, and mark each score known or guessed. (4) Total the scores; write next to the total how many of the six were guessed. (5) Record the sheet in the idea log with the date. Deliverable: a dated one-page scorecard per problem. Pitfalls: scoring the solution's elegance instead of the problem's quality; scoring intensity from your own irritation rather than the customer's; treating a total of guesses as a fact. A problem with a high total and four guessed scores is a problem to interview about, not a problem to build for.

DO THE MATH What a problem is worth Illustrative figures. A problem's economic weight is roughly its frequency, times its cost per occurrence, times the number of people who have it. Take the bookkeeper's shoebox: a small business spends about four hours a year sorting receipts before the tax deadline and its bookkeeper spends another three hours per client cleaning what arrives. At $45 an hour for the owner and $60 for the bookkeeper, one client's disorder costs about 4 × $45 + 3 × $60 = $360 a year. If the bookkeeper serves forty clients, the problem costs her practice about $7,200 a year in unbillable or underbilled time, plus the March stress that no invoice records. Across a region with 3,000 small businesses served by bookkeepers, the problem's weight is on the order of $1 million a year — enough to support a modest business if a solution can capture a fraction of it, and small enough that a founder should not promise a large one. Chapter 5 makes the estimate rigorous; the point here is that the arithmetic is possible on day one, with numbers a founder already has.

2.3 Trends and Timing

A problem that has existed for a century and remains unsolved is not automatically an opportunity; it may be unsolvable, or not worth solving, or already solved well enough. What makes a problem an opportunity now is usually a change — and the founder who can name the change has an answer to the question every investor, lender and co-founder asks: why now?

2.3.1 Five kinds of change

Changes that open opportunities come in five kinds, which the strategic analysis of Book A3 develops into the full PESTEL scan. Technology makes a solution possible or cheap that was impossible or costly: online payments, mobile devices, cloud software, AI assistants, cheaper sensors. Regulation creates obligations that businesses must meet and cannot meet alone — privacy law, accessibility rules, environmental reporting, food safety — and removes barriers that protected incumbents. Demographics shift who has the problem: an aging population, a wave of business owners retiring (a trend that Book M1 treats in full), newcomers, a generation that expects to buy everything online. Behaviour changes what people are willing to do: remote work, subscription habits, distrust of advertising, willingness to talk to a machine. Economics changes what people can afford and what they must save: interest rates, labor shortages, input costs, currency.

A trend is not an idea. "Everyone is aging" is a fact about the population; "owners of trades businesses aged 58 to 66 in the region have no successor and no plan, and their accountants have no service to offer them" is an opportunity statement that a trend made timely. The founder's task is to trace the trend down to a specific group, a specific problem and a specific reason the old solution stopped being good enough.

2.3.2 Adoption curves and the early market

Everett Rogers's research on how innovations spread describes a population that adopts in waves: a small group of innovators, a larger group of early adopters, then the early and late majorities, and finally the laggards, in proportions of roughly 2.5%, 13.5%, 34%, 34% and 16% of the eventual adopters (Rogers, 2003). Geoffrey Moore added the observation that matters most to a founder: between the early adopters and the early majority lies a chasm, because the two groups buy for different reasons and do not refer each other (Moore, 2014). The early adopters buy a change; the majority buys a proven solution to a known problem, with references.

For a new venture, the practical consequences are two. First, the first customers are early adopters almost by definition, and they should be sought where early adopters are — among the people who have already tried to solve the problem themselves, who built the spreadsheet, who complained in a forum. Second, the size of the eventual market is not the size of the early market, and a founder who sizes the venture on the majority before crossing the chasm builds for customers who will not arrive for years. Chapter 5 sizes both.

2.3.3 Reading the timing signal

Timing is one of the leading causes of failure in the post-mortem evidence of Chapter 1, and it is also one of the easiest causes to detect early, because a market that is not ready leaves a clear signal: no urgency. People agree the problem exists, find the solution interesting, and do nothing. Chapter 3's interviews are designed to hear that signal, and Chapter 9's tests are designed to measure it. A founder who hears "interesting" five times and "when can I have it?" zero times has learned something about timing, not only about the product.

QUÉBEC & CANADA NOTE Where trend and market data live (as of September 2026) Statistics Canada publishes the population and business data that turn a trend into a number: the Canadian Business Counts (businesses by industry, region and employee size, updated twice a year — the December 2024 counts are the current reference used in Book B1; Statistics Canada, 2025), the census profiles for demographics by region, the Survey of Digital Technology and Internet Use, and The Daily for business entry, exit and survival releases (Statistics Canada, 2026a). The Institut de la statistique du Québec publishes the Québec equivalents, including regional economic and demographic profiles (Institut de la statistique du Québec, 2026). BDC publishes free studies and guides on small-business trends and start-up practice. Innovation, Science and Economic Development Canada publishes the Key Small Business Statistics each year (Innovation, Science and Economic Development Canada, 2025). Sector associations publish member surveys that are often the only data on a niche. And search data — what people type into a search engine, available from the engines' own trend tools — is the cheapest signal of behavior, though it measures curiosity rather than demand. Contact and reference details for these sources are gathered in the Where to Get Help section of the master bibliography.

2.4 Jobs-to-Be-Done as an Opportunity Lens

A customer does not want a product; they want progress. Clayton Christensen and his colleagues framed this as the theory of jobs to be done: people "hire" a product or a service to make progress in a specific circumstance, and they "fire" it when something else makes the progress better (Christensen, Hall, Dillon, & Duncan, 2016). The theory matters for finding opportunities because it redirects attention from what people buy to what they are trying to accomplish — and the gap between the two is where new businesses fit.

DEFINITION Job-to-be-done A job-to-be-done is the progress a person is trying to make in a particular circumstance, described independently of any product. A job has a functional dimension (get the receipts to the accountant before the deadline), an emotional dimension (stop dreading March), and a social dimension (not look disorganized to the accountant). A product is a candidate hire for the job; it competes with every other candidate, including doing nothing. Naming the job, its circumstance and its three dimensions is the first step of Chapter 4's value proposition.

2.4.1 Jobs, circumstances and the alternatives already hired

The circumstance is what makes a job precise. "Feeding my family" is a category; "feeding two children on a weeknight when I get home at six with no plan and no energy" is a job in a circumstance, and it explains why a meal-kit subscription, a rotisserie chicken, a frozen lasagna and a neighbor's leftovers compete for the same hire. A founder who names the circumstance can list the alternatives the customer has already hired, which is exactly the list Chapter 8 needs for competitive analysis and Chapter 4 needs for differentiation.

Anthony Ulwick's version of the theory adds a useful discipline: describe the outcomes a customer wants from a job in a standard form — a direction, a measure, an object and a context, such as "minimize the time it takes to find last year's invoices when the auditor asks" — and then ask customers to rate each outcome for importance and for satisfaction with current solutions (Ulwick, 2005). Outcomes that are important and poorly satisfied are underserved; they are the opportunities. Outcomes that are unimportant and well satisfied are overserved; a cheaper, simpler offer that drops them is also an opportunity, as Section 2.5 explains.

2.4.2 Using the lens on your candidate problems

Take each problem from the scorecard and rewrite it as a job with a circumstance and three dimensions. Then list what the person hires today for that job, and what each hire does well and badly. Three patterns signal an opportunity. The job is done by an assembly — the customer combines three tools, a spreadsheet and a phone call to make the progress, which means no single hire fits. The job is done with an expensive hire — a professional, a full product, a subscription — for a simple version of the job, which means an overserved customer. And the job is not done at all — the customer has given up, which is the non-consumption that Section 2.5 treats as the largest market of all.

CASE IN POINT The bookkeeper who noticed the spreadsheet Fictional example; names and figures are invented. Mariam Haddad runs a bookkeeping practice with forty small-business clients in a mid-sized city. Every March she loses a week to receipts that arrive in bags, envelopes and photographs, and every year three or four clients ask her which app they should use, try one, and abandon it by June. Her first idea was the solution-first kind: a receipt-scanning app "that actually works". Rewriting the problem as a job changed it. The circumstance was not "receipts all year" but "the two weeks before the tax deadline, when the owner discovers the year's disorder at once." The functional job was to get the year's receipts to Mariam in a usable form; the emotional job was to stop dreading her call; the social job was not to look like a mess in front of the person who sees the numbers. The alternatives hired were a shoebox, a phone camera, an abandoned app and, for two clients, a part-time assistant. The overserved outcome was "capture every receipt in real time" — the apps all did it, and nobody wanted to. The underserved outcome was "get through March without shame or a week of work." Her opportunity brief, written that evening, proposed not an app but a quarterly receipt clinic: a two-hour appointment, in person or by video, where she and the client sorted the quarter together, with a fixed fee. Twelve clients said yes before she had a name for it. Chapter 3 tells what the interviews then revealed.

2.5 Market Gaps and Underserved Segments

A market gap is a place where a job is done badly, expensively or not at all for a group of people that existing suppliers have not organized themselves to serve. Gaps leave evidence, and a founder who knows what the evidence looks like can find them without a market study.

2.5.1 Non-consumption

The largest market for almost any new offer is the people who do not buy the category at all — not because they lack the job, but because the existing solutions are too expensive, too complicated, too far away or too intimidating for them. Clayton Christensen's early work on disruptive innovation showed how simple, affordable offers aimed at non-consumers repeatedly grew into markets that incumbents, busy serving their best customers, had ignored (Christensen, 1997). For a small venture, non-consumers are attractive for a reason beyond size: they compare the new offer with nothing, not with a polished incumbent, and they are grateful for a solution that a demanding customer would find crude. The evidence of non-consumption is the person who says "I just don't do that" about a job they clearly have.

2.5.2 Over-service

The mirror image is the customer who pays for a full product and uses a tenth of it. Small businesses that pay for enterprise software, households that pay for a cable package, clients who hire a professional for a routine task: each is overserved, and each will switch to an offer that does the tenth they use, simply and cheaply. The evidence of over-service is the customer who cannot say what most of the product does, or who complains about the price of a product they "have to have".

2.5.3 Workarounds and the spreadsheet signal

When people build their own solution, they are signalling three things at once: the job matters enough to spend time on, no existing offer fits, and they have already designed the first version of the product for you. The spreadsheet is the most common workaround signal in business markets; in consumer markets it is the group chat, the shared document, the routine assembled from three apps, the neighbor who "always knows a guy". The evidence is easy to collect: ask, in the interview of Chapter 3, "what do you do about it today?" and listen for the assembly.

2.5.4 Complaints, reviews and the questions people ask

Public complaints are a free research department. Reviews of existing products, forum threads, questions asked in professional groups, support tickets in a business the founder works for: each records a job the current hires do badly, in the customer's own words. The founder's task is to read enough of them to see a pattern rather than an anecdote — Chapter 3 gives the thresholds — and to note the exact words, which Chapter 4 will reuse in the value proposition.

Table 2.2 gathers the sources of opportunity described in this chapter and, for each one, the evidence it leaves and the question that surfaces it. Read it as a field guide: a founder looking for opportunities does not need inspiration but a list of places to look and a question to ask in each. The last column names the chapter where the evidence is then tested.

Table 2.2 — Sources of opportunity and the evidence they leave

2.6 Shaping the Opportunity

An idea in a founder's head is a bundle of assumptions. Shaping it means writing it down in a form that separates what is known from what is guessed, that another person could read and challenge, and that leads directly to the next test.

2.6.1 The opportunity statement

The shortest useful form is one sentence, and it is written before any mention of a solution: "[Specific people] in [circumstance] struggle to [job] because [obstacle]; today they [workaround], which costs them [cost]." Mariam's version read: "Owners of one-to-five-person businesses, in the two weeks before the tax deadline, struggle to get a year of receipts to their bookkeeper in usable form because they never had a routine; today they send a shoebox or a phone full of photos, which costs them a week of stress and costs their bookkeeper three unbilled hours per client." Every clause of that sentence is a hypothesis that Chapter 3 can test, and none of it mentions an app.

The solution comes second, and it comes as a class rather than as a design: "We believe that [a class of solution] would let them [outcome], and that they would [pay / switch / refer] because [reason]." Writing the solution as a class — a service, a tool, a subscription, a marketplace, a course — keeps the founder from committing to a design before the problem is confirmed, and it makes the alternatives visible: Mariam's class was "a periodic guided sorting service", of which the receipt clinic was one design among several.

2.6.2 The one-page opportunity brief

METHOD The opportunity brief When to use it: for each of the two or three problems that survive the scorecard, before the discovery sprint. Steps: (1) Problem statement — the one sentence of Section 2.6.1, in the customer's words. (2) Who — the specific group, with an estimate of how many there are in the region you can reach, and three named examples. (3) Evidence so far — what you know from your own experience, from the people you have spoken to, from data; each item marked fact or impression. (4) Alternatives hired today — the workarounds and the products, with what each does well and badly. (5) Why now — the change that makes the problem timely, named specifically. (6) Our access — what you have that makes this cheaper for you than for others. (7) Solution class and first hypotheses — the class of solution, and the three riskiest assumptions in the form of Chapter 1. (8) Next test — the first thing you will do to test the riskiest assumption, with a date. Deliverable: one page, dated, in the idea log. Pitfalls: writing the brief around the solution; leaving section 3 as a list of impressions; skipping the named examples in section 2 (if you cannot name three people, reachability is a guess); writing "why now" as a slogan rather than a change.

2.6.3 Brief, canvas and hypotheses

The brief is the seed of everything that follows. Its problem statement becomes the customer profile of Chapter 4; its "who" becomes the segment and the beachhead of Chapter 5; its alternatives become the competitive map of Chapter 8; its solution class and hypotheses become the first entries of the assumption map of Chapter 10; and its "next test" is the first task of the discovery sprint of Chapter 11. A founder who keeps the brief current — revised after each round of interviews, each version dated — has a record of how the opportunity was shaped, which is what a lender, an investor or a co-founder will later ask to see.

WATCH OUT Solutions in search of a problem The most common failure of shaping is to write the brief backwards: to start from a product the founder wants to build and then describe a problem that would justify it. The tell-tale signs are a problem statement that names the product's features ("struggle to scan receipts in real time"), a "who" defined as "anyone who…", evidence that consists of the founder's own enthusiasm, and a "why now" that is really "why me". The consequence is the Rideline case of Chapter 1: a year and $62,000 spent confirming that a problem existed, followed by the discovery that the job was different. Shape from the problem; hold the solution loosely; let Chapter 3 decide.

2.7 Screening Many Ideas

A founder with one idea is at risk of marrying it. A founder with twenty ideas, scored, is in a position to choose — and choosing well is the whole point of this chapter. The screen is a funnel, and Figure 2.1 draws it: many ideas enter, the problems behind them are scored, the best few are written as briefs, and one enters the discovery sprint. The kill criteria on the side of the figure are the reasons an idea leaves at each stage; they are listed so that an idea is set aside for a written reason rather than for fatigue.

Figure 2.1 — The opportunity screen. Twenty or more ideas are collected in the idea log; the problems behind them are scored with Table 2.1; about three are written as one-page opportunity briefs; one enters the discovery sprint of Chapter 3. The kill criteria remove ideas at each stage, and killed ideas stay in the log with their reason. Original figure.

2.7.1 Collecting ideas without judging them

The first stage is generous by design. For two weeks, the founder writes every problem noticed — at work, at home, in conversations, in complaints read online — as one line in the idea log, without scoring or solving. Twenty is a minimum; forty is common once the habit is installed. Solutions that come to mind are written on the same line, in parentheses, and ignored. The log is kept in one place (a note in La Machine, a document, a spreadsheet) with a date on every entry, because the order in which problems were noticed is itself information about the founder's access.

2.7.2 Scoring and choosing three

The second stage applies the scorecard of Table 2.1 to every problem in the log. Most scores will be guesses at this stage, and that is expected; the purpose is to rank, not to conclude. The eight or so problems with the highest totals get a second look, during which the founder rewrites each one as a job with a circumstance (Section 2.4) and lists the alternatives hired today. Three of them become opportunity briefs. Three, not one, because the discovery sprint of Chapter 3 will kill or transform at least one of them, and a founder with a second brief ready loses a week rather than a quarter.

2.7.3 The kill criteria

An idea leaves the screen for one of seven reasons, and the reason is written in the log next to the idea. The problem is rare or mild, so that nobody will change a habit to solve it. Nobody can be found who has it, so that the founder is describing a problem they imagine rather than one they observed. The people who have it will not pay, because they have no budget, or the workaround is free and good enough. No segment can be reached at a cost the model can bear, because the people are scattered and anonymous. The solution needs capital, skills or permits that the founder cannot obtain in the time available. The law forbids it, or the timing is clearly not now. Or the founder, honestly, would not want to run this business for five years — a criterion that is not a weakness but the most reliable predictor of whether the venture survives its first bad year.

2.7.4 The screen in the three situations

Table 2.3 shows how the screen changes with the founder's situation. It compares the solo founder, the founding team and the established business on four dimensions: where the ideas come from, who scores them, how many briefs are written and who decides which one enters discovery. The pattern is that the screen becomes more formal as more people are involved, not because the criteria change, but because a written reason is the only way several people can set an idea aside without resentment.

Table 2.3 — The opportunity screen in the three situations

WHO DOES THIS Owning the screen In a solo venture the founder owns the idea log, the scoring and the choice, and the risk is a screen that is really a rationalization of the first idea; the remedy is the outsider who reads the scorecards and asks which scores were guessed. In a founding team, one member keeps the log and the briefs, all members score, and the team agrees before scoring how the choice will be made, so that the decision is not a negotiation between attachments. In an established business, the person who owns discovery reads the customer records that employees do not think of as ideas — the complaints, the workarounds sales reps improvise, the questions service answers every week — and presents scorecards rather than pitches at the gate. In all three, the person who wrote a brief is the wrong person to score it alone.

Costly Mistakes

1. Starting from a solution and describing a problem to fit it. The brief names the product's features as the customer's struggle, the "who" is "anyone who…", and the evidence is enthusiasm. The consequence is a discovery sprint that confirms a problem exists and misses that the job is different — a quarter lost, and often a year.
2. Scoring the idea's size instead of the problem's quality. A big market with a mild, rare problem beats a small market with a weekly, expensive one on every spreadsheet and loses to it in every real business. The consequence is a venture that must create demand before it can sell, at a cost a new business cannot bear.
3. Treating a trend as an idea. "Everyone is aging" or "AI changes everything" is visible to a thousand founders at once and gives none of them an advantage. The consequence is a crowded, undifferentiated entry, or a market that is not ready — the timing failure of the post-mortems.
4. Keeping one idea. With a single candidate, every interview becomes a defense and the first kill criterion ends the venture. The consequence is either a quarter of denial or a founder with nothing to do next; three briefs cost two more evenings and remove both.
5. Setting ideas aside without a written reason. An idea killed by fatigue comes back six months later and consumes another quarter, because nobody remembers why it was dropped. The consequence is repeated discovery of the same dead end.
6. Skipping the named examples. A brief whose "who" has no three named people has guessed reachability and access, the two criteria that most often kill a venture after the product is built. The consequence is a validated problem with no way to reach the people who have it.

Key Takeaways

• Opportunities come from what the founder already has — problems lived, skills held, access owned and trends noticed — and the research on founders shows that most successful ventures started that way, with modest means and quick adaptation rather than a grand plan.
• An opportunity is a problem that a reachable group has often and intensely enough to pay for, at a time when a solution has become possible, from a position that gives the founder access. An idea is a candidate; evidence makes it an opportunity.
• Fall in love with the problem, not the solution. Write the problem first, in the customer's words, and treat every solution — including your first — as a hypothesis.
• Score problems on six criteria — frequency, intensity, cost of the workaround, whether the payer feels the pain, reachability and your access — and mark which scores are guesses; a high total of guesses is a reason to interview, not to build.
• Timing is a named change, not a slogan. Technology, regulation, demographics, behavior or economics must have changed something specific for a specific group, and a market that is not ready signals it by the absence of urgency.
• The job-to-be-done lens reveals the gap between what people buy and the progress they seek; the largest opportunities are in non-consumption, over-service and the workarounds people build for themselves.
• Shape the idea into a one-page brief — problem, who, evidence, alternatives, why now, access, hypotheses, next test — and keep three briefs alive, because the discovery sprint will kill or transform at least one.
• Set ideas aside for a written reason. The kill criteria, applied and recorded, are the cheapest lessons a venture will ever learn.

Your Action Plan

The plan below runs over two weeks and produces the three opportunity briefs that Chapter 3 needs. Each step names the deliverable, the time it takes and the criterion by which you will know it is done. Steps 1 and 2 collect and score; steps 3 to 5 shape and choose; step 6 records the decision so that the choice survives the first difficult interview.

In La Machine

La Machine does not find opportunities, but it holds the log, the briefs and the decision in a way that survives the founder's memory, and it offers one tool that shortens the search for reachable people.

• The Business Explorer (business/marketing/business-explorer), opened with Explore Businesses on CRM → Companies, searches five sources — Google Maps, OpenStreetMap, a Canadian business directory, a consultants directory and a service-pros directory — by name, industry and city, and gathers the results in a staging table that can be exported or converted into company prospects. For the founder screening ideas, it answers the reachability question of the scorecard in minutes: which bookkeeping practices, plumbing contractors or dental clinics exist in the region, and where, within a cap of twenty results per search on the map sources and two hundred on the directories, whose database must be installed. The page is open to the administrator and reseller roles, records the source of every result, and displays the attribution that open map data requires.
• Notes (/presse-papier) keep the idea log: one note per problem, colored by status, attachable to a project or a task. Links (/liens) keep the data sources of the Québec and Canada Note one click away.
• Projects (/projets) group the tasks of each opportunity brief, and whoever administers the platform loads the project, since no page creates one; the scorecard and the job statement are attached as document links to the brief's first task, because the Projects page does not display a project's documents. Tasks (/taches) carry the "next test" of section 8 of the brief as a task with a verification criterion and a date, which is set when the task is loaded, since a task created through the assistant is born without one.
• Arbitrages (/arbitrages) records the choice of the first brief and the reasons the others wait — the entry that the action plan's step 6 requires.

What La Machine does not do: it has no trend-monitoring or search-volume tool, and no integrated market data — those come from the sources of the Québec and Canada Note and from the tools of Chapter 13. It does not score problems or write briefs; the scorecard and the brief are the founder's work, held in the platform. Market research as a discipline is the subject of Book B2 (Market Research and Competitive Intelligence). The two paths by which prospects later enter the CRM, the prospect warehouse and the Business Explorer, are covered in Book N4 (CRM, Contacts and Prospect Data), and the workflows that automate a venture's logistics are covered in Book N3 (Workflows, Automation and Human Approval).

Self-Check

1. Name the four sources of opportunity described in this chapter, and say which one gives a founder the least advantage over other founders, and why.
2. What are the six criteria of the problem scorecard, and what does a high total made mostly of guessed scores tell you to do?
3. State the difference between a painkiller and a vitamin, and why a first venture should usually be a painkiller.
4. Rewrite "people need a better receipt app" as a job with a circumstance and three dimensions.
5. Give three of the seven kill criteria, and explain why a killed idea stays in the log.

Answers. (1) Problems lived, skills held, access owned and trends noticed; a trend gives the least advantage because it is visible to everyone at once, and a founder who has only noticed it has no access others lack (Section 2.1.1). (2) Frequency, intensity, cost of the workaround, whether the payer feels the pain, reachability and your access; a high total of guesses tells you to interview about the problem, not to build for it (Section 2.2.3). (3) A painkiller removes a problem the customer already feels and already tries to solve; a vitamin promises an improvement the customer does not yet feel the lack of, so the founder must create demand before selling, which a new venture cannot afford (Section 2.2.2). (4) For example: "Owners of very small businesses, in the two weeks before the tax deadline, need to get a year's receipts to their bookkeeper in a usable form (functional), to stop dreading the call (emotional), and not to look disorganized in front of the person who sees their numbers (social)" (Sections 2.4 and 2.6.1). (5) Any three of: the problem is rare or mild; nobody can be found who has it; the people who have it will not pay; no segment can be reached at a bearable cost; the solution needs capital, skills or permits the founder cannot obtain; the law forbids it or the timing is not now; the founder would not want to run it for five years. A killed idea stays in the log with its reason so that it is not rediscovered and re-investigated six months later (Section 2.7.3).

Chapter References

Bhidé, A. V. (2000). The origin and evolution of new businesses. Oxford University Press.

Christensen, C. M. (1997). The innovator's dilemma: When new technologies cause great firms to fail. Harvard Business School Press.

Christensen, C. M., Hall, T., Dillon, K., & Duncan, D. S. (2016). Competing against luck: The story of innovation and customer choice. HarperBusiness.

Innovation, Science and Economic Development Canada. (2025). Key small business statistics 2025. Government of Canada. Retrieved September 7, 2026, from https://ised-isde.canada.ca/site/sme-research-statistics/en/key-small-business-statistics/key-small-business-statistics-2025

Institut de la statistique du Québec. (2026). Statistiques et publications. Gouvernement du Québec. Retrieved September 7, 2026, from https://statistique.quebec.ca/

Moore, G. A. (2014). Crossing the chasm: Marketing and selling disruptive products to mainstream customers (3rd ed.). HarperBusiness.

Rogers, E. M. (2003). Diffusion of innovations (5th ed.). Free Press.

Sarasvathy, S. D. (2001). Causation and effectuation: Toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academy of Management Review, 26(2), 243–263.

Shane, S. (2003). A general theory of entrepreneurship: The individual-opportunity nexus. Edward Elgar.

Statistics Canada. (2025). Canadian Business Counts, with employees, December 2024 (Table 33-10-0764-01). Retrieved September 7, 2026, from https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3310076401

Statistics Canada. (2026a). The Daily. Retrieved September 7, 2026, from https://www150.statcan.gc.ca/n1/dai-quo/index-eng.htm

Ulwick, A. W. (2005). What customers want: Using outcome-driven innovation to create breakthrough products and services. McGraw-Hill.

Chapter 3 — Customer Discovery and Validation

You have described your idea to twenty people and eighteen of them said it was a great idea. You are now the owner of eighteen compliments and zero facts. This chapter replaces the compliments with evidence: what the people you want to serve actually do, what they have tried, what it costs them, and what they will commit to — before you build anything.

What You'll Be Able to Do

• Tell evidence from opinion, using the evidence ladder — what customers say, what they do, what they pay — and weigh each rung correctly in a decision.
• Conduct twelve customer interviews with a script that asks about past behavior rather than future intentions, and avoid the five mistakes that make interviews worthless.
• Design and run a short survey at the right moment, size its sample honestly and read its results without fooling yourself.
• Collect behavioral evidence from observation, workarounds, search data, reviews and the records a business already holds.
• Obtain commitments — time, reputation, money — through pre-orders, deposits and letters of intent, within the rules that apply to them.
• Keep a validation log and synthesize evidence into patterns with thresholds, so that the brief of Chapter 2 is revised on facts.
• Run research ethically and lawfully: consent, purpose, honesty and the handling of the personal information you collect.

Why It Matters

Every method that follows in this book — the value proposition, the market sizing, the canvas, the numbers, the tests — consumes what this chapter produces. A value proposition written from twelve interviews describes real pains in real words; one written from the founder's imagination describes the founder. A market sized from observed behavior is a plan; one sized from a survey of friends is a wish. For the founder, discovery is the cheapest insurance the venture will ever buy. For a team or an established business, it is the only argument that ends a meeting: not "I think" but "seven of twelve said, unprompted, that they lose a day a month to this, and four gave a deposit."

In This Chapter

• 3.1 Evidence over Opinion
• 3.2 Customer Interviews
• 3.3 Surveys and Quantitative Signals
• 3.4 Observation and Existing Data
• 3.5 Pre-Sales, Pre-Orders and Letters of Intent
• 3.6 Synthesizing Evidence
• 3.7 Ethics and Compliance in Research

3.1 Evidence over Opinion

Steve Blank's instruction to founders was short: there are no facts inside the building, so get outside. His customer development process runs in four steps — customer discovery, in which the founder tests the problem and the solution hypotheses with customers; customer validation, in which a repeatable sales process is proved; customer creation; and company building — and it insists that the first two are done by the founders themselves, in person, before scaling anything (Blank, 2013; Blank & Dorf, 2012). This chapter is the first of those steps, done at the scale of a small venture.

DEFINITION Customer discovery Customer discovery is the structured search for evidence about who has the problem, how they experience it, what they do about it today and what they would commit to change — conducted by the founders, in conversation with and observation of the people the venture intends to serve, before the offer is built. Its output is not a list of features but a set of confirmed or refuted hypotheses and a revised opportunity brief.

DEFINITION Validation Validation is the confirmation of a hypothesis by evidence strong enough to justify the next commitment. A hypothesis about a problem is validated by a pattern of behavior across enough people; a hypothesis about willingness to pay is validated by payment or a binding commitment; a hypothesis about a channel is validated by people found and reached through it at a bearable cost. Validation is always relative to a decision: the evidence that justifies a second week of interviews is much weaker than the evidence that justifies a year of building.

3.1.1 The evidence ladder

Not all evidence weighs the same. Figure 3.1 arranges it on three rungs. On the lowest rung is what people say: interview answers and survey responses, valuable as a starting point and weak alone, because people misremember, please the interviewer and predict their own behavior badly. On the middle rung is what they do: the workaround they built, the time they spend, the sign-up they completed, the way they used a prototype — behavior that costs effort and reveals what words hide. On the top rung is what they pay: a pre-order, a deposit, a signed letter of intent, a paid pilot, a renewal, a referral. Cindy Alvarez put it as a hierarchy of currencies: the strongest validation is money, then reputation — a referral, an introduction, a public endorsement — then time (Alvarez, 2014).

Figure 3.1 — The evidence ladder. What customers say (interviews, surveys) is a signal to explore; what they do (observation, workarounds, sign-ups, use of a prototype) is a pattern to confirm; what they pay (pre-orders, deposits, letters of intent, pilots, renewals) is a reason to commit. Twelve interviews find the pattern, a survey measures how common it is, a commitment proves that it matters. Original figure.

3.1.2 What counts as evidence

Chapter 1 defined decision-grade evidence as behavioral, specific and sufficient. The ladder adds the practical corollary: a decision should rest on the highest rung that the decision's cost justifies. Choosing which problem to interview about can rest on what people say. Choosing which segment to serve first should rest on what they do. Choosing to build, to quit a job or to borrow should rest on what they pay. A founder who has climbed the ladder in that order has a model that has met customers three times before it met a developer.

KEY CONCEPT Compliments are not evidence, and neither are opinions about the future The two most abundant outputs of an unstructured customer conversation are compliments ("great idea", "I'd definitely use that") and predictions ("I'd pay $50 a month for it"). Both are free for the customer to give and both are worthless as evidence, because they cost nothing and commit to nothing. Rob Fitzpatrick's rule is that the only reliable information a conversation yields is about the customer's life — what they did, when, how often, what it cost, what they tried — not about your idea (Fitzpatrick, 2013). Every script in this chapter is built to ask about the past and to listen for behavior, and every synthesis method counts behaviors, not enthusiasm.

3.2 Customer Interviews

The customer interview is the founder's primary instrument, and it is a skill: most first interviews are bad, most tenth interviews are good, and the difference is the script, the rules and the habit of writing down what was said rather than what was hoped.

3.2.1 Whom to interview

Interview the people described in the "who" of the opportunity brief, and start with those who show the early-adopter signals of Chapter 2: they have the problem now, they have tried to solve it, they have spent time or money on a workaround, and they can be reached through a channel you already have. Interview across the decision unit when the buyer, the user and the payer are different people — a bookkeeper, her client and the client's spouse who keeps the receipts are three interviews, not one. Include people who have the problem and have not tried to solve it; their reasons are the timing signal. And after the first six, deliberately recruit two or three people who do not fit the brief's "who", because their answers draw the boundary of the segment.

3.2.2 How many

A founder often asks how many interviews are enough. The research on qualitative interviews gives a usable answer: in a study designed to measure it, the basic themes were present after six interviews and nearly all of them (94%) had appeared after twelve, with little new after that — a phenomenon the authors called saturation (Guest, Bunce, & Johnson, 2006). The number depends on how homogeneous the group is; a tightly defined segment saturates faster than "small businesses". Twelve per segment is the working rule of this book, with a checkpoint after six.

DO THE MATH How many interviews, and what a pattern is Illustrative figures, built on the saturation finding above. Plan twelve interviews per segment, in two rounds of six. After the first six, list every theme heard; if the fifth and sixth interviews added no new theme, the segment is homogeneous and the second round will confirm rather than discover. Set thresholds before the second round: a theme heard unprompted from 6 or more of 12 is a pattern strong enough to write into the value proposition; a theme heard from 3 to 5 is a signal worth a specific question in the next round or a survey; a theme heard from 2 or fewer is an anecdote, recorded and not acted on. Two segments therefore cost about twenty-four interviews of thirty to forty minutes each — roughly 16 hours of conversation plus 8 hours of recruiting and 6 of synthesis, about 30 hours, or five weeks of the six-hour slot. That is the largest single investment of the discovery quarter, and the one that every later chapter repays.

3.2.3 The script

The script asks about the past, in the customer's setting, and it is short enough to remember. It opens with the circumstance: "Tell me about the last time you had to [job]." It follows the story: what happened next, who was involved, how long it took, what it cost. It asks about frequency: "How often does that happen?" It asks about the workaround: "What do you do about it today? What have you tried? Why did you stop?" It asks about cost, in the customer's terms: hours, dollars, stress, relationships. It asks about the decision unit: "Who else cares about this? Who would have to agree?" It asks about the alternatives: "If [workaround] disappeared tomorrow, what would you do?" It closes with the two questions that produce the highest-rung evidence a conversation can yield: "Who else should I talk to?" — a referral is reputation — and, only after the problem is confirmed, "If a solution existed that did [outcome], what would you need to see before trying it?" The solution itself is not pitched (Constable, 2014); if the customer asks what you are building, the honest answer is that you are trying to understand the problem first.

METHOD The interview script and the five rules When to use it: every discovery interview, first round and second. Steps: (1) Recruit with a one-line purpose — "I'm researching how [people like you] handle [job]; thirty minutes, no sales pitch" — and a scheduling link. (2) Open with the last time the problem happened and follow the story. (3) Ask frequency, workaround, cost, decision unit, alternatives, in that order, with silence after each answer. (4) Write down verbatim quotes, marked with quotation marks, and note the behaviors separately from the opinions. (5) Close with "who else should I talk to?" and, if the problem is confirmed, "what would you need to see?" (6) Within an hour, file the notes in the validation log with the date, the person's segment and the hypotheses touched. The five rules, adapted from Fitzpatrick (2013): talk about their life, not your idea; ask about specifics in the past, not opinions about the future; talk less and listen more — a good interview is 80% the customer's voice; never pitch, and if you slip, mark everything said afterwards as contaminated; and end every conversation by asking for a commitment or a referral, because a conversation that ends in a compliment has taught you nothing. Deliverable: one filed note per interview, with at least three verbatim quotes and one behavior. Pitfalls: interviewing only friends; leading questions ("wouldn't it be great if…"); hypotheticals; one interview per role and calling it a pattern; scheduling twelve in one week and losing the ability to adjust the script after six.

3.2.4 Running the interview

Thirty to forty minutes is enough; an hour is a favor that few people grant twice. Two people are better than one — one asks, one writes — and a founding team should rotate roles so that both learn to listen. Recording is useful and requires consent stated at the start (Section 3.7); if the person hesitates, take notes instead. Interview in the customer's setting when possible — the shop, the office, the kitchen — because the setting shows the workaround and the interruptions that the customer forgets to mention. And write the notes within the hour, before memory rewrites them into what you wanted to hear.

3.2.5 The mistakes that make interviews worthless

Five mistakes account for most useless interviews. Pitching turns the interview into a sales call and the customer into a polite audience; every answer after the pitch is about your idea, not their life. Leading questions — "do you find it frustrating that…" — supply the answer; ask "how do you feel about…" or better, "what happened?". Hypotheticals — "would you use…", "would you pay…" — collect predictions that the customer cannot make; ask what they did the last time and what they paid for the workaround. Accepting compliments as data inflates the log with enthusiasm that costs nothing; count behaviors. And interviewing only people you know produces a sample that likes you; the second round must include strangers found through the segment's own channels.

WATCH OUT The interview that confirms everything If twelve interviews confirm every hypothesis of the brief, the interviews were wrong, not the brief. Real discovery kills something: a feature nobody mentioned, a segment that turned out not to have the problem, a price the customer laughed at, a decision-maker the founder had not imagined. The most common cause of a clean sweep is a script that pitched and then asked for agreement. The second most common is a sample made of friends. A founder who finishes a round with nothing to change should re-read the notes for behaviors only, ignoring every opinion, and count again.

CASE IN POINT Twelve interviews that killed a feature and found a segment Fictional example; names and figures are invented. Mariam Haddad (Chapter 2) recruited twelve owners of small businesses — six current clients and six strangers reached through a local business association — for thirty-minute conversations about "the last time you got your receipts to your bookkeeper." She did not mention the receipt clinic. Nine of twelve described the same two weeks of March, unprompted; the pattern was real. But the interviews changed three things. First, the feature she had assumed — a shared folder where clients would drop photos all year — was mentioned by nobody as something they wanted and by four as something they had already abandoned: "I'm not going to photograph a parking receipt in the rain." Second, the segment split. Owners with revenue under about $150,000 CAD shrugged at the problem ("it's one bad week"), while owners above $300,000, with staff and no assistant, described it as a crisis that delayed their tax filing and, in two cases, cost them a penalty. Third, the emotional job was stronger than she had guessed: five used the word "ashamed" or "embarrassed" without being asked. When she then described the quarterly clinic — only after the problem had been fully discussed — seven of the twelve asked when it would start, and four accepted, on the spot, to pay a $150 deposit for the first session. The validation log recorded nine problem confirmations, one feature invalidated, one segment boundary discovered, five emotional quotes, seven expressions of intent and four deposits. The brief was rewritten that weekend around owners above $300,000 with staff; the shared folder never appeared again.

3.3 Surveys and Quantitative Signals

Surveys are the most overused and least understood instrument of discovery. Used at the wrong moment they produce confident nonsense; used at the right moment they answer the one question interviews cannot: how common is the pattern?

3.3.1 When a survey helps

A survey helps after interviews have found a pattern and defined the words, and its job is to measure frequency and importance across a larger group: what share of the segment has the problem monthly, what share uses which workaround, which of three outcomes matters most. A survey does not help before interviews, because the founder does not yet know which questions to ask or which words the customer uses, and the questions will measure the founder's assumptions. It does not help for willingness to pay, because stated intentions are the bottom rung of the ladder; Chapters 7 and 9 give better instruments.

3.3.2 Designing the questions

The research on survey design converges on a few rules that a founder can apply without training (Dillman, Smyth, & Christian, 2014). Keep it short: eight to twelve questions, under five minutes, and say so at the start. Ask one thing per question. Use the customer's words from the interviews. Prefer closed questions with exhaustive, non-overlapping options for anything you will count, and one open question at the end for anything you have not imagined. Avoid leading wording, double negatives and questions that assume the problem. Ask about behavior and frequency ("in the last three months, how many times…") before importance and satisfaction ("how important is…; how satisfied are you with how you handle it today…"). Ask the classification questions — role, size, sector — at the end, and only those you will use to cut the results. Test the survey on three people from the segment before sending it, and fix what they misread.

3.3.3 Samples, response rates and bias

Two numbers decide what a survey can say: how many people answered, and who they were. The Do the Math box gives the arithmetic of sample size; the harder problem is bias. People who answer a survey about a problem are more likely to have the problem than those who ignore it, so a survey overstates prevalence unless the response rate is high or the sample was drawn from a list that represents the segment. A founder should therefore report a survey as "of the 84 owners who answered, 61% said…", never as "61% of owners". And a survey sent to a founder's own network measures the network, not the market; the second round of interviews and the survey should both reach strangers through the segment's channels — an association, a supplier's list, a trade publication, a community.

DO THE MATH How many responses a survey needs Illustrative figures. The margin of error of a proportion measured on a random sample is approximately z × √(p(1 − p) / n), where z is 1.96 for a 95% confidence level, p is the proportion (use 0.5 for the widest margin) and n is the number of responses. Solving for n gives n = z² × p(1 − p) / e². For a margin of ±10 percentage points, n = 1.96² × 0.25 / 0.10² = 96 responses; for ±5 points, n = 384. A discovery survey that receives 96 usable answers from the segment can say "between 51% and 71% of respondents report the problem monthly" — precise enough to size a market in Chapter 5, and honest about its limits. With 30 answers the margin is about ±18 points, which still separates "most" from "few" and is often all a founder needs. The formula assumes a random sample; a self-selected sample adds a bias that no sample size removes, which is why the sentence always begins with "of those who answered".

Table 3.1 compares the research methods of this chapter. For each method it states what the method measures, which rung of the evidence ladder it reaches, its typical cost in a founder's hours and dollars, the moment in the discovery quarter when it is used, and its main weakness. Read it as a sequencing guide: the methods near the top find and explain, the methods near the bottom confirm and commit, and the discovery quarter uses them in roughly that order.

Table 3.1 — Research methods compared

3.4 Observation and Existing Data

The middle rung of the ladder is the least used and, for the hours it costs, the most revealing. People cannot describe what they do not notice they do, and a business's own records hold evidence that nobody has read as evidence.

3.4.1 Observation in the customer's setting

Watching a customer perform the job in their own environment — a method that design practitioners call contextual inquiry — shows the interruptions, the improvisations and the workarounds that interviews miss (Beyer & Holtzblatt, 1998). Two hours in a bookkeeper's office in March teach more about the receipt problem than five interviews, because the shoebox is on the desk and the phone rings with the client who cannot find the invoice. Observation is arranged as a favor, with consent, and the founder watches and asks only clarifying questions; the notes record what happened in what order, how long each step took, and what the person said while doing it.

3.4.2 Workarounds and artefacts

The workaround is evidence a customer has already produced. Ask to see it: the spreadsheet, the checklist taped to the wall, the folder structure, the group chat, the form they built. Each artefact records the customer's own design of the solution — the fields they thought mattered, the steps they could not automate, the point at which they gave up. Photograph it with permission and file it in the log; Chapter 4's value map is often a cleaned-up version of what the customer already built.

3.4.3 Existing data

A founder can read a great deal before the first interview. Search data from the search engines' trend tools shows how many people look for the problem, in which words, in which months and regions; it measures curiosity, not demand, and it is free. Reviews of existing products, in the two-star and three-star range especially, record what current hires do badly. Forums and professional groups record the questions people ask each other when no product answers them. Support and sales records in a business the founder works for or serves record the complaints and requests of real customers, week after week. Public statistics — the sources of Chapter 2's Québec and Canada Note — size the population that has the job. None of this replaces conversation, and all of it makes conversation sharper: a founder who has read forty reviews knows which words to listen for.

3.4.4 The business that already has customers

An established business investigating a new offer has an advantage it usually ignores: years of records. The service tickets, the requests that sales reps improvised answers to, the products returned, the questions asked at the counter, the clients who left — each is behavioral evidence about jobs the current offer does badly. The intrapreneur's first week of discovery should be spent reading those records with the question of Chapter 2 in mind: which job is being done by an assembly, by an expensive hire, or not at all?

3.5 Pre-Sales, Pre-Orders and Letters of Intent

The top rung of the ladder is commitment, and the strongest commitment is money. Asking for it before the offer exists feels presumptuous to most founders and is, in fact, the most respectful test there is: it asks the customer to answer with the only currency that cannot be faked by politeness.

3.5.1 The commitment ladder

Commitments come in three currencies, and a founder should climb them in order (Alvarez, 2014). Time is the weakest: a second meeting, an hour of observation, a review of a prototype. Reputation is stronger: a referral to a colleague, an introduction to a decision-maker, an agreement to be named as a pilot customer, a public endorsement. Money is strongest: a deposit, a pre-order, a paid pilot, a signed purchase commitment. Every interview should end by asking for the next rung the conversation justifies; a customer who confirmed the problem and described its cost can be asked for a referral, and a customer who asked "when can I have it?" can be asked for a deposit.

3.5.2 Pre-orders and deposits

A pre-order is a sale of something that does not yet exist, and it is lawful and common — with conditions. The offer must be described truthfully: what will be delivered, when, at what price, and what happens if delivery is late or the venture stops. Under the Competition Act, a representation to the public that is false or misleading in a material respect is prohibited whether or not a product exists (s. 74.01), so a pre-order page that promises features the founder has not decided to build is a legal risk, not a marketing tactic. Deposits should be refundable and held apart from operating money, so that a stopped venture can return them. And when the buyers are consumers rather than businesses, the rules on distance contracts apply, which the Québec and Canada Note summarizes.

THE RULE Pre-selling honestly (Canada, as of September 2026) Under the Competition Act, R.S.C. 1985, c. C-34, no person may, for the purpose of promoting a product or a business interest, make a representation to the public that is false or misleading in a material respect (s. 74.01(1)(a)); the general impression the representation conveys, as well as its literal meaning, is taken into account (s. 74.01(6)). In plain terms: a pre-order page, a pitch or a letter of intent must describe what will actually be delivered, by when and for how much, and must not imply that the product exists, works or is used by others if it does not, does not or is not. State the stage plainly ("in development; first deliveries planned for March; full refund if we do not deliver"), keep every claim in a file with its evidence (Chapter 4), and treat a pre-order as a promise you can keep. The rule applies to businesses of every size and to representations made to business buyers as well as to consumers.

QUÉBEC & CANADA NOTE Consumer pre-orders are distance contracts (Québec, as of September 2026) When the buyer is a consumer — a natural person buying for personal purposes — and the sale is concluded without the parties being in each other's presence, it is a distance contract under the Consumer Protection Act, CQLR c P-40.1 (s. 54.1). The Act, in its consolidated version of April 7, 2026, imposes obligations that a pre-order page must meet. The merchant may not collect a partial or full payment before performing its principal obligation unless the consumer may request a chargeback of the payment under the Act or a regulation (s. 54.3) — in practice, the consumer's chargeback right against a credit card issuer (ss. 54.14–54.16) is what makes an advance payment by credit card permissible; taking a consumer's deposit by another means should be checked with a lawyer or with the Office de la protection du consommateur, whose guide to distance contracts sets out the consumer's cancellation rights (Office de la protection du consommateur, 2026). Before the contract, the merchant must disclose its name, address and contact details, a detailed description of the goods or services, an itemized list of prices and additional charges, the total, the currency, the delivery date and mode, and the conditions of cancellation, return, exchange and refund (s. 54.4), and must give the consumer an express opportunity to accept or decline and to correct errors (s. 54.5). A copy of the contract must be sent within 15 days (s. 54.7). The consumer may cancel within seven days of receiving the copy if these obligations were not met (s. 54.8), and at any time before performance if the principal obligation is not performed within 30 days after the date specified — or within 30 days of the contract when no date was specified (s. 54.9); after a cancellation, the merchant must refund all sums within 15 days (s. 54.13). For a founder, the practical reading is simple: a consumer pre-order needs a page that discloses everything in section 54.4, a stated delivery date, payment by a means that preserves the chargeback right, and the capacity to refund within 15 days. Business-to-business pre-orders and deposits are not covered by this Act; they are governed by the contract itself (→ see Book H1, Contracts and Commercial Agreements).

3.5.3 Letters of intent

In business markets, the instrument between a conversation and a contract is the letter of intent: a short, usually non-binding document in which a prospective customer states that, if the offer is delivered as described, they intend to buy a stated quantity at a stated price or price range within a stated period, subject to stated conditions. Its value is reputational rather than legal — a buyer who signs has put their name to the intent and will be reluctant to disown it — and its strength rises with its specificity: a quantity, a price, a date and a named signatory. A letter of intent that says "we would be interested in learning more" is a compliment on letterhead. A founder should draft the letter, keep it to one page, state clearly whether it binds anyone (usually not, except for confidentiality), and ask for it at the end of the second conversation, when the problem, the outcome and the price range have been discussed. Lenders and investors read a file of specific letters of intent as evidence; Book A5 explains how they weigh it.

3.5.4 What a commitment does not prove

Four deposits prove that four people expect value; they do not prove that the venture can deliver it at a profit, that the segment is large, or that the price is right. The commitment test validates desirability at the top rung; viability and feasibility still need the numbers of Chapter 7 and the tests of Chapter 9. And a commitment obtained through the founder's personal relationship — a friend, a former employer — carries a weight discount that the validation log should record honestly.

3.6 Synthesizing Evidence

Evidence that is not written down, dated and sorted does not exist. The validation log is the instrument that turns twenty conversations into a revised brief, and the synthesis session is the ritual that keeps the founder honest.

3.6.1 The validation log

Table 3.2 lists the columns of the validation log. Each row is one observation — one thing one person said, did or paid — and the columns record where it came from, which hypothesis it touches, how strong it is and what it implies. The log is kept in one place from the first interview to the last test, and it is the single most valuable document of the discovery quarter: it is what a co-founder reads to join, what a lender reads to trust, and what the founder reads six months later to remember why the model looks the way it does.

Table 3.2 — The columns of the validation log

3.6.2 Thresholds and patterns

The thresholds set in the Do the Math box of Section 3.2 — a pattern at six of twelve, a signal at three to five, an anecdote below — are set before the second round, written in the log's header, and applied mechanically. They protect the founder from the two opposite errors: building on an anecdote because it was vivid, and dismissing a pattern because it was unwelcome. Patterns are counted on behaviors and unprompted statements; prompted agreement ("would you say that…?") is counted separately and weighted low.

3.6.3 The evidence board

For each hypothesis of the brief, the evidence board is a simple three-column view: evidence for, evidence against, and what is still unknown, each entry pointing to a row of the log. A hypothesis with a full "for" column and an empty "against" column after twelve interviews is either validated or was never really tested; a hypothesis with entries on both sides needs a decision about which side the behaviors favor. Chapter 10 turns the board into the assumption map and the decision to persevere, pivot or stop.

METHOD Synthesis in ninety minutes When to use it: after each round of six interviews, and after any test. Steps: (1) Print or lay out every verbatim quote and every behavior from the round, one per card, without the person's name. (2) Sort the cards into clusters by what they are about — the circumstance, the cost, the workaround, the decision unit, the outcome wanted — and name each cluster in the customers' words. (3) Count, for each cluster, how many different people it came from, and mark it pattern, signal or anecdote against the thresholds. (4) For each hypothesis of the brief, fill the evidence board from the clusters: for, against, unknown. (5) Rewrite the opportunity brief — the problem statement, the "who", the evidence section — using only patterns, and date the new version. (6) Write the three questions the next round must answer and adjust the script. Deliverable: the clusters with counts, the updated evidence board, the dated brief, the revised script. Pitfalls: clustering by the founder's categories instead of the customers' words; counting statements instead of people; letting one vivid interview name a cluster; skipping step 5 because "nothing changed" — something always changes, and the version history proves it.

WHO DOES THIS Discovery in the three situations In a solo venture the founder conducts every interview and does the synthesis alone, which is why the thresholds must be written down in advance: they are the only outsider in the room. A friend who reads the log and challenges the clusters is worth an hour a week. In a founding team, one member owns the log and the script, all members interview in pairs and rotate roles, and synthesis is done together on the cards, not from memory; disagreements about a cluster are settled by counting people. In an established business, the intrapreneur interviews personally and does not delegate the first twelve to a sales representative, who will pitch, or to an agency, which will report opinions; the business's records (Section 3.4.4) are read before the interviews, and the synthesis is presented at the gate as clusters with counts and quotes, never as a slide of conclusions.

3.7 Ethics and Compliance in Research

Discovery collects personal information — names, contact details, opinions, recordings, sometimes financial details — from people who trusted the founder enough to talk. The rules that protect them are also good research practice, because people who understand what will happen to their words speak more freely.

THE RULE Consent and purpose when collecting personal information for research (Québec and Canada, as of September 2026) In Québec, the Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25) requires a person who collects personal information to inform the individual, at the time of collection, of the purposes for which it is collected, the means of collection, the individual's rights of access and rectification, and their right to withdraw consent (s. 8); it limits use to the purposes for which the information was collected unless the person consents (s. 12); and it requires that consent be manifest, free, enlightened and given for specific purposes, for the time needed to achieve them (s. 14). Federally, the Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5 applies the principles of consent, limiting collection and limiting use. In plain terms for a founder: say at the start of every interview who you are, why you are asking, what you will keep (notes, a recording, a name), whom you will share it with, and how the person can withdraw; ask before recording; keep the notes without the name where a code will do; delete the recordings once the synthesis is done; do not reuse the contact details to send marketing without a separate, unbundled consent, because a promotional message to an interviewee is a commercial electronic message under CASL s. 6 that requires its own consent, identification and unsubscribe mechanism. The obligations of Law 25 on a written privacy policy and a person in charge of personal information apply as soon as the venture holds personal information in the course of an enterprise; Book H2 (Privacy and Personal Information) gives the full regime and Book B1, Chapter 10, the marketing consent rules.

3.7.1 Honesty about the venture

Tell interviewees the stage the venture is at. "We are studying the problem and have not built anything" is a sentence that lowers the customer's guard and raises the quality of the evidence, because there is nothing to be polite about. If a prototype is shown, say that it is a prototype. If a pre-order is asked for, say what happens to the money if the venture stops. The founder who is honest about uncertainty collects better evidence and, later, better customers.

3.7.2 Incentives, confidentiality and the NDA reflex

A small incentive for an interview — a gift card, a coffee, a copy of the findings — is common and should be disclosed as such; it does not bias a conversation about past behavior, though it may bias a hypothetical, which is one more reason not to ask hypotheticals. Confidentiality runs in both directions: the founder protects the interviewee's information, and the interviewee is told whether the findings will be shared and in what form. The reflex to ask interviewees to sign a non-disclosure agreement before describing the idea should be resisted: experienced customers and advisors refuse them, they signal that the founder values the idea over the evidence, and an idea that can be stolen by a description had no defensible advantage anyway (Chapter 8 discusses what actually protects a venture). Book H5 (Intellectual Property) covers the cases where confidentiality genuinely matters — a technical invention before filing, a supplier's pricing — and how to handle them.

3.7.3 Research with vulnerable or regulated groups

Interviewing minors, patients, people in financial distress or employees of a business the founder works for raises additional duties — parental consent, professional secrecy, employment rules — that this chapter cannot cover. The safe rule is to recognize the situation and to ask a professional before proceeding; the Where to Get Help section of the master bibliography names the regulators.

Costly Mistakes

1. Pitching in the interview. Every answer after the pitch concerns the founder's idea rather than the customer's life, and the log fills with polite enthusiasm. The consequence is a value proposition written from compliments and a segment that turns out not to buy — the Rideline outcome of Chapter 1, discovered a year late.
2. Surveying before interviewing. The survey measures the founder's assumptions in the founder's words, with a self-selected sample, and returns confident percentages about the wrong question. The consequence is a market sized on nonsense and a plan that inherits it.
3. Counting statements instead of people, and prompted agreement as unprompted evidence. One vivid interviewee who said the same thing five times becomes a "pattern", and twelve nods to a leading question become "everyone agrees". The consequence is building on an anecdote — typically a feature nobody uses and a quarter of engineering.
4. Stopping at "says". A dozen confirmations that the problem exists are taken as validation of the offer, without a single behavior observed or a dollar committed. The consequence is discovering at launch that people who agree they have a problem will not change what they do to solve it.
5. Taking consumer pre-orders without the distance-contract rules. A page that takes a deposit by e-transfer, states no delivery date and lists no cancellation conditions violates the Consumer Protection Act on several points at once. The consequence is cancellations, refunds on demand, complaints to the regulator and a reputation problem before the first delivery.
6. Collecting personal information without stating the purpose. Names, emails and recordings gathered for research and later used for marketing breach Law 25 and CASL together. The consequence is exposure to administrative penalties that can reach into the millions for a company (Book B1, Chapter 10), and, more commonly, the loss of the goodwill of the very people who helped.

Key Takeaways

• Evidence sits on a ladder — what customers say, what they do, what they pay — and a decision should rest on the highest rung its cost justifies: interviews to choose a problem, behavior to choose a segment, commitments to build.
• Compliments and predictions are not evidence. The only reliable information from a conversation is about the customer's past behavior: the last time, the frequency, the workaround, the cost, the people involved.
• Twelve interviews per segment, in two rounds of six, find the pattern. Themes present after six are nearly complete after twelve; the thresholds — pattern at six of twelve, signal at three to five — are written before the second round and applied mechanically.
• The script asks about the past and ends with a commitment. Talk about their life, ask about specifics, listen more than you speak, never pitch, and close with a referral or the next rung.
• Surveys measure how common a known pattern is; they do not discover it. Design them from the interviews' words, keep them short, size the sample honestly and always say "of those who answered".
• Observation, workarounds and existing records are the cheapest behavioral evidence, and an established business already holds years of it in tickets, returns and complaints.
• Commitments have three currencies — time, reputation, money — and pre-selling is lawful when it is honest: truthful representations under the Competition Act, and, for consumers in Québec, the disclosure, payment, delivery and refund rules of the Consumer Protection Act.
• The validation log is the venture's most valuable document. One observation per row, dated, sourced, weighted and tied to a hypothesis; a ninety-minute synthesis after each round rewrites the brief on patterns alone.
• Consent and purpose are stated at the start of every conversation, recordings are deleted after synthesis, and research contacts are never turned into a marketing list without a separate consent.

Your Action Plan

The plan below runs over about five weeks of the six-hour slot and produces the evidence on which Chapters 4 and 5 are written. Each step names its deliverable, its duration and the criterion by which it is verified. Steps 1 to 4 conduct and synthesize the interviews; step 5 adds behavioral evidence; step 6 climbs to commitment; step 7 secures the log.

In La Machine

La Machine holds the people, the forms and the records of discovery, and it enforces the consent rules that research must respect.

• Forms (business/marketing/forms, responses at business/marketing/forms/responses) carry the discovery survey and the interview sign-up, and the responses are kept with the form, dated, ready for the log. As read on September 15, 2026, a checkbox added to a form records an answer, not a consent: only a lead magnet's own consent box, unticked and optional, documents a consent at the moment of capture, and its label becomes the proof sentence, so it names who will write and about what — the unbundled consent that Law 25 requires. A sign-up that will receive messages is therefore taken through a lead magnet, or its consent is documented afterwards on the contact's consent page (→ see Book N5).
• Contacts (business/marketing/contacts) record each interviewee with a source, a tag for the segment and five per-channel opt-in flags — email, text message, push, voice and direct message — that stay off until a consent is documented, with its proof, on the contact's consent page; the contact's record holds the interview notes as notes and dated activities, so that the second round finds the first round's quotes. A research contact who did not consent to marketing therefore cannot receive a campaign, and the do-not-reach list (business/marketing/do-not-reach) is kept for the person who asks not to be contacted: it records the email address or the telephone number with its reason, for every business of the account, and only an administrator removes an entry.
• Meetings (productivity/meetings) schedule the interviews and the observation sessions, and each session is logged as an activity of the type Meeting or Call on the interviewee's record, under Productivity → Tasks (productivity/tasks); such an activity can be grouped in a productivity project, which has no link with the plan's projects, so its description names the project of the brief.
• Projects and tasks (/projets, /taches) hold the discovery sprint of Chapter 11: recruiting, interviewing, synthesis and commitment tasks, each with its verification criterion — "six notes filed, each with three quotes and one behaviour".

What La Machine does not do: it does not transcribe or analyze interviews, cluster quotes or compute survey margins — the synthesis is done on cards and in a spreadsheet, or with the tools of Chapter 13 — and it does not provide search data, reviews or public statistics. The validation log itself lives in a document or a spreadsheet attached to the project; Book N4 (CRM, Contacts and Prospect Data) shows how contacts, consent and forms fit together, and Book B2 (Market Research and Competitive Intelligence) develops the research methods that this chapter only introduces.

Self-Check

1. Name the three rungs of the evidence ladder and give one example of evidence on each.
2. Why are compliments and predictions worthless as evidence, and what should a founder ask about instead?
3. How many interviews per segment does this chapter recommend, in what structure, and what thresholds define a pattern, a signal and an anecdote?
4. When does a survey help in discovery, and what must a founder always add before reporting its percentages?
5. Name three obligations that a consumer pre-order page in Québec must meet under the Consumer Protection Act, and the rule of the Competition Act that applies to every pre-sale.

Answers. (1) Says — an interview answer about the last time the problem occurred; does — a workaround the customer built, or a sign-up; pays — a deposit, a signed letter of intent or a paid pilot (Section 3.1.1). (2) They cost the customer nothing and commit to nothing; a founder should ask about the customer's past behavior — the last time, the frequency, the workaround, the cost, the people involved (Section 3.1.2 and 3.2.3). (3) Twelve per segment in two rounds of six, with synthesis between; a pattern is a theme heard unprompted from six or more of twelve different people, a signal from three to five, an anecdote from two or fewer (Section 3.2.2). (4) After interviews have found the pattern and the words, to measure how common it is; every result is reported as "of those who answered", because a self-selected sample overstates prevalence (Section 3.3). (5) Any three of: disclosure of the merchant's identity, the description, the itemized prices, the total, the delivery date and the cancellation and refund conditions before the contract (s. 54.4); an express opportunity to accept, decline and correct errors (s. 54.5); no advance payment unless the consumer has a chargeback right (s. 54.3); a copy of the contract within 15 days (s. 54.7); refund within 15 days of cancellation (s. 54.13). Every pre-sale must avoid representations that are false or misleading in a material respect under Competition Act s. 74.01 (Section 3.5.2).

Chapter References

An Act to promote the efficiency and adaptability of the Canadian economy by regulating certain activities that discourage reliance on electronic means of carrying out commercial activities, S.C. 2010, c. 23 (Canada's Anti-Spam Legislation), s. 6. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/E-1.6/

Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25), ss. 8, 12, 14. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1

Competition Act, R.S.C. 1985, c. C-34, s. 74.01. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-34/

Consumer Protection Act, CQLR c P-40.1, ss. 54.1–54.16 (consolidated to April 7, 2026). LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-40.1

Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/P-8.6/

Alvarez, C. (2014). Lean customer development: Building products your customers will buy. O'Reilly Media.

Beyer, H., & Holtzblatt, K. (1998). Contextual design: Defining customer-centered systems. Morgan Kaufmann.

Blank, S. (2013). The four steps to the epiphany: Successful strategies for products that win (5th ed.). K&S Ranch.

Blank, S., & Dorf, B. (2012). The startup owner's manual: The step-by-step guide for building a great company. K&S Ranch.

Constable, G. (2014). Talking to humans: Success starts with understanding your customers. Giff Constable.

Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, phone, mail, and mixed-mode surveys: The tailored design method (4th ed.). Wiley.

Fitzpatrick, R. (2013). The mom test: How to talk to customers and learn if your business is a good idea when everyone is lying to you. Founder Centric.

Guest, G., Bunce, A., & Johnson, L. (2006). How many interviews are enough? An experiment with data saturation and variability. Field Methods, 18(1), 59–82.

Office de la protection du consommateur. (2026). Online purchases: Cancelling a distance contract. Gouvernement du Québec. Retrieved September 7, 2026, from https://www.opc.gouv.qc.ca/en/consumer/topic/purchase/online-purchase/cancelling/

Chapter 4 — The Value Proposition

A founder describes the product for four minutes — the features, the technology, the roadmap — and the prospective customer nods and asks, "So what does it do for me?" The founder has a product. The customer is asking for a value proposition. This chapter builds it from the evidence of Chapter 3, so that the answer to "so what?" is in the customer's words, ranked by what matters to them, and ready to be tested.

What You'll Be Able to Do

• Define value and perceived value, and explain why value is always relative to the alternative the customer would otherwise choose.
• Fill a customer profile — jobs, pains and gains, ranked — from interview evidence rather than imagination.
• Fill a value map — products and services, pain relievers, gain creators — and connect each item to the profile it addresses.
• Judge fit at three levels: problem–solution fit, product–market fit and business-model fit, with the evidence each one requires.
• Differentiate in ways customers value, and pass the "only" test against the real alternatives.
• Write a value proposition statement in one of four formats, and test it with five strangers.
• Make claims you can prove before you have customers, within the rules on truthful representations.

Why It Matters

The value proposition is the sentence that every other part of the venture serves. It decides what the product must do and what it may leave out; it decides which segment the venture serves first, because a proposition is written for someone; it is the message that the website, the sales conversation and the pre-order page will carry; and it is the promise against which customers will judge the venture after they pay. For the founder, a tested proposition turns "we do many things" into "we do this one thing for these people, better than what they do today." For a team, it is the agreement that ends debates about features: if a feature does not relieve a ranked pain or create a ranked gain, it waits.

In This Chapter

• 4.1 Value, Perceived Value and the Exchange
• 4.2 Customer Jobs, Pains and Gains
• 4.3 Products, Pain Relievers and Gain Creators
• 4.4 Fit
• 4.5 Differentiation That Customers Value
• 4.6 Writing the Value Proposition Statement
• 4.7 Claims and Early Proof

4.1 Value, Perceived Value and the Exchange

The word value is used loosely in business, and the looseness hides the mechanism that makes a venture work. This section fixes the meaning before the canvas uses it.

4.1.1 Value is what the customer gets minus what they give, as they perceive it

Valarie Zeithaml, reviewing how consumers actually judge value, defined it as the customer's overall assessment of the utility of a product, based on their perception of what is received and what is given (Zeithaml, 1988). Three words in that definition do the work. Received includes everything the customer gets — the outcome, the time saved, the risk removed, the feeling, the status — not only the product. Given includes everything they pay — the price, but also the time to learn, the effort to switch, the risk of being wrong, the discomfort of change. And perceived means that the customer's judgment, not the founder's calculation, decides; a benefit the customer does not notice does not exist, and a cost the customer fears counts even if the founder knows it is small.

The consequence for a new venture is that value has two sides to design. A founder can raise value by increasing what is received — a better outcome, a faster result — or by lowering what is given — a lower price, a simpler start, a guarantee that removes the risk. Most first ventures over-invest in the first and ignore the second, and then wonder why a clearly better product does not sell: the customer perceives the switching cost, and the founder does not.

4.1.2 Value is relative to the alternative

No customer compares an offer with nothing. They compare it with what they do today — the workaround, the incumbent, the professional they hire, or doing nothing at all — and value is the difference between the two, as they see it. James Anderson, James Narus and Wouter van Rossum, studying value propositions in business markets, found that the propositions that actually sell are not the lists of all benefits that suppliers like to write, but the ones that name the one or two points of difference that matter most to the customer against the next best alternative — what they called a resonating focus (Anderson, Narus, & van Rossum, 2006). Chapter 2 collected the alternatives in the interviews; Chapter 8 maps them; this chapter writes the proposition against them.

DEFINITION Value proposition A value proposition is the set of outcomes, pain relievers and gains that a specific customer segment receives from an offer, stated relative to the alternatives that segment uses today, in terms the segment itself uses. It is not a slogan, a mission statement or a feature list. It is written for one segment at a time — a second segment gets a second proposition — and it is a hypothesis until customers have confirmed, by behavior or payment, that they value what it says.

KEY CONCEPT Value is relative to the alternative, and the alternative is often "nothing" The question "what is our value proposition?" is unanswerable until the founder has answered "compared with what?". For most first customers, the alternative is a workaround or nothing at all, which means the proposition competes with the customer's tolerance of the problem, not with a rival product. That is easier in one way — no incumbent to displace — and harder in another: the founder must make the cost of the status quo visible before the customer will consider paying to remove it. The pains and their costs, collected in Chapter 3, are the material for that.

4.1.3 The 9× problem

John Gourville, studying why clearly superior products fail to be adopted, combined two findings from behavioral economics. Customers weigh what they might lose by switching about three times more heavily than what they might gain, and developers of a new product, who have lived with its benefits, overvalue them by about three times relative to a customer who has not. The mismatch multiplies to a factor of about nine between what the developer thinks the customer will perceive and what the customer actually perceives (Gourville, 2006). For a founder, the lesson is arithmetic rather than rhetoric: a proposition must offer a gain that is large in the customer's perception and a switching cost that is small, and the founder's own enthusiasm is a poor instrument for measuring either. The Do the Math box in Section 4.4 turns this into a test.

4.2 Customer Jobs, Pains and Gains

Alexander Osterwalder and his colleagues organized the customer side of a value proposition into three lists — jobs, pains and gains — that together form the customer profile, and the offer side into three matching lists — products and services, pain relievers, gain creators — that form the value map; fit is the match between the two (Osterwalder, Pigneur, Bernarda, & Smith, 2014). Figure 4.1 draws the two panels with example items from the receipt-clinic case; this section fills the profile, Section 4.3 fills the map.

Figure 4.1 — Customer profile and value map. The customer profile (right) lists the jobs the segment is trying to get done, the pains it experiences and the gains it seeks, each ranked. The value map (left) lists the products and services offered, the pain relievers and the gain creators. Fit is the match between the two on the top-ranked items. Original figure, adapted from the concept described by Osterwalder, Pigneur, Bernarda and Smith (2014); example items from the fictional receipt-clinic case.

4.2.1 Jobs

Jobs are what Chapter 2 defined: the progress the customer is trying to make in a circumstance, with a functional, an emotional and a social dimension. The profile lists them from the interviews — in the customers' words, not the founder's — and ranks them by importance to the customer. Two rules keep the list honest. A job is written independently of any solution ("get a year's receipts to the bookkeeper in usable form", not "upload receipts to an app"). And a job is included only if it appeared as a pattern or a strong signal in the validation log; a job the founder believes in but nobody mentioned goes on the list in italics, as a hypothesis, or not at all.

4.2.2 Pains

Pains are what makes the job hard, risky or unpleasant: undesired outcomes ("I miss the deadline"), obstacles ("I can't find last year's invoices when the auditor asks"), and risks ("if I get it wrong there's a penalty"). Each pain is ranked by severity — how much it costs, how often, how strongly the customer described it — and the ranking uses the counts of the synthesis: a pain named unprompted by nine of twelve outranks one named by three, regardless of the founder's view. The most useful pains carry a number the customer gave: hours, dollars, a penalty, a lost customer.

4.2.3 Gains

Gains are the outcomes and benefits the customer wants: required gains without which the offer is useless (the receipts reach the bookkeeper), expected gains that any decent solution provides (it is not humiliating), desired gains that would delight if present (a quarterly picture of where the money went), and unexpected gains the customer would not think to ask for (a bookkeeper who knows the business well enough to answer a question in March). Gains are ranked by relevance, and the same rule applies: only gains that appeared in the evidence, in the customer's words, are listed as facts.

4.2.4 One profile per segment

A profile describes one segment. When Chapter 3's interviews split the segment — as Mariam's did at about $300,000 of revenue — the founder writes two profiles and discovers that they differ on the top-ranked items: the smaller owners' top pain was "one bad week", the larger owners' was "a penalty and a delayed filing". Two profiles that differ on the top items require two value maps and, usually, two propositions, and the beachhead decision of Chapter 5 chooses which one comes first.

4.3 Products, Pain Relievers and Gain Creators

The value map is the offer side, and it is written only after the profile, because its purpose is to answer the profile item by item.

4.3.1 Products and services

List what the offer consists of — the product, the service, the digital tool, the support, the training, the guarantee — plainly and completely, including the parts the founder considers secondary. A customer buying a receipt clinic is also buying the envelope kit that arrives before it, the reminder that schedules it and the year-end summary that follows; each of those is a product or service that costs something to deliver (Chapter 7) and relieves something (below).

4.3.2 Pain relievers

For each ranked pain of the profile, write how the offer removes, reduces or makes bearable that pain — specifically, not generically. "Saves time" is not a pain reliever; "the year's receipts are sorted in four two-hour sessions instead of one lost week in March" is. A pain reliever that does not correspond to a ranked pain is a feature looking for a reason, and it is either dropped or listed as a hypothesis. The strongest maps relieve the top two or three pains completely and ignore the rest; a map that relieves twelve pains a little is a product nobody can describe.

4.3.3 Gain creators

For each ranked gain, write how the offer produces it. The discipline is the same: specific, tied to a ranked gain, and honest about the level — a required gain must be delivered every time, a desired gain is a differentiator, an unexpected gain is a story the customer will tell. Gain creators are where a small venture often finds its edge, because incumbents built for the required gains have stopped noticing the desired ones.

4.3.4 The "so what" chain

A practical test for every item on the map is to ask "so what?" until the answer is an outcome the customer named. "We provide pre-sorted envelopes" — so what? — "receipts are grouped by category as they arrive" — so what? — "the quarterly session takes two hours instead of four" — so what? — "no lost week in March, which nine of twelve owners named as their top pain." The chain ends at a profile item or it does not end at all, in which case the item is a feature, not value.

WATCH OUT Feature lists disguised as value The most common draft of a value proposition is a list of the product's features with the word "benefit" attached: "cloud-based", "AI-powered", "easy to use", "all-in-one". None of those is a pain relieved or a gain created; each is a claim about the product that the customer must translate into their own outcome, and most will not bother. The consequence is a website and a pitch that describe the founder's work rather than the customer's progress, and a conversion rate that stays near zero however good the product is. The test is the "so what" chain: every line of the proposition must end at a ranked item of the profile, in the customer's words, or be cut.

CASE IN POINT Two value maps for one product Fictional example; names and figures are invented. Norvale Compost collects organic waste from restaurants in a mid-sized city and turns it into compost sold to landscapers. Its founders, two former kitchen managers, wrote one value proposition — "affordable organic waste collection that keeps you compliant" — and sold nothing for two months. Twelve interviews (Chapter 3) revealed two profiles behind the word "restaurant". The owner's jobs were to control costs and avoid fines; the top pains were the price of waste collection, the risk of being caught non-compliant with organic-waste rules that were tightening in the region, and the reputational cost of a rat complaint; the desired gain was a certificate to show customers. The kitchen manager's jobs were to run a clean, fast service; the top pains were the smell and the fruit flies of a bin that sat for a week, the space the bin took, and staff who refused to sort; the required gain was a bin that was emptied before it stank. The same collection service, on two value maps, relieved different pains: for the owner, a fixed monthly price below the current contract and a compliance certificate; for the manager, twice-weekly pickup in summer, a sealed bin that fit under the counter, and a one-page sorting poster in three languages. Two propositions, two conversations, two decision-makers — and the sale, it turned out, was made in the kitchen and signed in the office. Norvale's first ten contracts came in the six weeks after the maps were separated.

4.4 Fit

Fit is the match between the value map and the customer profile, and it exists in degrees and levels. A founder who can say which level of fit the venture has reached, and on what evidence, has a language for the whole discovery quarter.

DEFINITION Fit (problem–solution, product–market, business-model) Problem–solution fit is reached when there is evidence that the segment cares about the jobs, pains and gains the offer addresses and that the offer, as designed, addresses them — evidence from interviews and observation, before the offer exists in usable form. Product–market fit is reached when customers use, pay for and keep the offer at rates that justify growing it — evidence from pilots, sales and retention, once a usable version exists. Business-model fit is reached when the model that delivers the offer is viable — the unit economics and the costs of Chapter 7 work at a scale worth the effort. The three are reached in order, and a venture that has the first is not entitled to claim the second.

4.4.1 Problem–solution fit

The evidence for problem–solution fit is the evidence of Chapter 3 applied to the profile and the map: the top-ranked jobs, pains and gains are patterns in the log, not the founder's beliefs; the pain relievers and gain creators address those top-ranked items; and customers, shown the map or a description of the offer after the problem has been discussed, respond with behavior — a referral, a request to be first, a deposit. Mariam's seven "when can I have it?" and four deposits are problem–solution fit. Twelve people agreeing that receipts are a nuisance is not.

4.4.2 Product–market fit

Product–market fit, a term that Marc Andreessen popularized as "being in a good market with a product that can satisfy that market", is recognized by its symptoms: customers buy as fast as the venture can serve them, usage grows, word spreads, and the founder's problem becomes delivery rather than demand (Andreessen, 2007). Because those symptoms are easier to see in hindsight than in the moment, practitioners use a leading indicator: a short survey of current users asking how they would feel if they could no longer use the product, with the share answering "very disappointed" as the measure — a threshold of about 40% is treated as the signal that the product has become a must-have for a core group (Ellis & Brown, 2017). The test is meaningful only once real customers have used a real offer for long enough to miss it, which places it in Chapter 9's pilots, not here.

4.4.3 Business-model fit

A proposition that customers love can still describe a business that cannot exist: the pain relievers cost more to deliver than the customer will pay, the gain creators require a resource the founder cannot obtain, the channel to reach the segment eats the margin. Business-model fit is the subject of Chapters 6 and 7, and it is mentioned here for one reason: the value map should be written with an eye on what each item costs, so that the proposition does not promise what the model cannot afford.

DO THE MATH The switching test Illustrative figures. A customer switches when the gain they perceive exceeds the cost they perceive, by a margin large enough to overcome the weight they give to losses. Take the owner in Norvale's case. The perceived gain of switching: a monthly collection price of $180 instead of the current $240 ($60 a month, $720 a year), plus a compliance certificate the owner values at perhaps $500 a year in avoided worry and marketing use, for a perceived gain of about $1,220 a year. The perceived cost: an hour to cancel the current contract and set up the new one ($60 at the owner's rate), the staff retraining the manager estimates at three hours ($75), the fear that a new supplier fails during a summer weekend — which the owner mentally prices at, say, $400 — and the general cost of change, for a perceived cost of about $535, plus the loss-aversion multiplier of roughly three on the risk of failure. Net perceived value: $1,220 − ($135 + 3 × $400) = −$115. The offer loses, however good it is on paper, because the fear of a failed pickup dominates. The map, revised: a guaranteed pickup with a $100 credit for any missed one, and a first month free. The fear falls to near zero, the perceived gain rises, and the same offer now clears the test by about $1,000. That is what the switching test is for — to find the item on the map that the founder's enthusiasm hid.

4.5 Differentiation That Customers Value

A value proposition that a competitor could sign is not a proposition; it is a category description. Differentiation is the part of the map that the alternatives do not offer — and it is valuable only if customers value it.

4.5.1 Three ways to state value, and the one that sells

Anderson, Narus and van Rossum observed three kinds of value proposition in business markets. The all-benefits proposition lists everything the offer does; it is easy to write and it asks the customer to do the work of finding what matters. The favorable points of difference proposition lists what the offer does that the next best alternative does not; it is better, and it still assumes that every difference matters. The resonating focus proposition names the one or two points of difference — and sometimes one point of parity — that matter most to the customer, with proof; it is the hardest to write, because it requires knowing the customer's ranking and the alternative's performance, and it is the one that sells (Anderson et al., 2006). Chapter 3's rankings and Chapter 8's alternatives grid are what make a resonating focus possible for a venture that does not yet have customers.

4.5.2 Four tests of a differentiator

A point of difference earns a place in the proposition when it passes four tests. It is valued: it relieves a top-ranked pain or creates a top-ranked gain, in the evidence. It is believable: a customer who hears it does not laugh, and the founder can show why it is true. It is provable: there is, or will be, evidence a customer can check (Section 4.7). And it is defensible enough to survive the competitor's response for the time the venture needs to establish itself — a question Chapter 8 develops as unfair advantage. A differentiator that fails the first test is decoration; one that fails the last is a head start, which is not nothing, but which should be treated as temporary.

4.5.3 The "only" test

The shortest test of differentiation is a sentence that begins with "only": "We are the only [category] that [does X] for [segment]." If the sentence cannot be completed truthfully, the venture has parity, not difference, and must compete on price or on access. If it can, the founder checks that X is on the customer's top-ranked list; "we are the only receipt service that uses colour-coded envelopes" is true and worthless, while "we are the only bookkeeping practice in the region that sorts the year with you, every quarter, at a fixed fee" is true and relieves the top pain. The alternatives grid of Chapter 8 verifies the "only".

4.5.4 Differentiation for a small venture

Small ventures rarely win on the product alone. Their most credible differentiators are usually specificity (built for one segment's circumstance, where the incumbents serve everyone), access (the founder's relationship, location, language or expertise), the experience around the product (speed, guarantees, a human who answers), and the model itself (a fixed fee where the alternative bills hourly, a subscription where the alternative sells a project). Each of these can be written as a point of difference that customers value, and each is cheaper to build than a better product.

4.6 Writing the Value Proposition Statement

The profile, the map and the differentiators condense into a statement of one to three sentences that a founder, a co-founder and a first employee can all recite and that a customer can repeat after hearing it once. Several formats work; Table 4.1 gives four, with an example of each from the receipt-clinic case, the situation in which each is most useful and its main risk. The formats differ in what they lead with — the customer, the outcome, the circumstance or a number — and a founder should draft the proposition in at least two formats and test both, because the format that feels natural to the founder is often not the one the customer repeats.

Table 4.1 — Four formats for a value proposition statement

4.6.1 Rules for the statement

Six rules carry over from the positioning work of Book B1, Chapter 5, and apply to a venture without customers. The statement uses the customer's words from the log, not the founder's. It names a real alternative, so that "unlike" has content. It excludes someone — a proposition for everyone is a proposition for no one — and the exclusion is a feature, not an apology. Every claim in it is provable or is marked as a hypothesis. It is short enough to be repeated after one hearing. And it is stable enough to be true for a year, because the website, the pre-order page and the sales conversation will all carry it.

METHOD Writing and testing the statement in two hours When to use it: after the profile and the map are filled from patterns, and before any page or pitch is written. Steps: (1) Copy the top-ranked job, the top two pains and the top gain from the profile, in the customer's exact words. (2) Copy the pain relievers and gain creators that address them from the map. (3) Write the alternative the segment uses today, from the interviews. (4) Draft the statement in two formats of Table 4.1. (5) Run the four tests on each draft: the repeat test — read it once to five people from the segment, preferably strangers, and ask them to say back what the offer is and for whom; four of five should repeat the segment and the outcome; the "only" test of Section 4.5.3; the "so what" test of Section 4.3.4 on every clause; and the "prove it" test — for every claim, name the evidence you have or would need. (6) Keep the format that passed the repeat test; record both drafts, the test results and the date in the log. Deliverable: one statement, dated, with its test record and its list of claims to prove. Pitfalls: testing on friends; testing by asking "do you like it?" instead of "say it back"; keeping the founder's favorite format after the customer repeated the other one; writing the statement before the profile, and then fitting the profile to it.

4.7 Claims and Early Proof

A proposition is a set of claims, and a venture that has not yet delivered anything must be especially careful about what it claims and how it proves it. The rules that govern claims apply from the first pre-order page, and the discipline of proof is what turns a proposition into a trustworthy one.

THE RULE Truthful claims before launch (Canada, as of September 2026) Under the Competition Act, R.S.C. 1985, c. C-34, it is prohibited, for the purpose of promoting a product or a business interest, to make a representation to the public that is false or misleading in a material respect (s. 74.01(1)(a)), or to make a representation about the performance, efficacy or length of life of a product that is not based on an adequate and proper test, the proof of which lies on the person making the representation (s. 74.01(1)(b)); the general impression conveyed counts as well as the literal meaning (s. 74.01(6)). In plain terms: a proposition may not claim a result the venture has not achieved, describe as existing what is planned, or state a number that cannot be substantiated; a performance claim ("filed early for the first time", "cuts the March week to four two-hour sessions") requires a test or a record that supports it before it is made. The Competition Bureau enforces these provisions, and the administrative penalties for a company can reach the greater of $10 million (more for a repeat), three times the benefit obtained, or 3% of annual worldwide gross revenues (Book B1, Chapter 10 gives the regime in full). The rule applies to representations to business buyers as well as to consumers, and it applies to a founder's pitch and pre-order page exactly as it applies to an established company's advertising.

4.7.1 What you can say before you have customers

A venture without customers has more provable material than founders think, and the honest presentation of it is itself a differentiator. Evidence from discovery can be stated as what it is: "in twelve interviews with owners of businesses above $300,000, nine described losing a week every March" — a true statement about research, not a performance claim. The founder's credentials and experience — years in the trade, a license, a professional designation — are provable and relevant. The mechanism can be described plainly: what the offer does and how, without promising results the mechanism has not yet produced. Pilot results, once pilots exist (Chapter 9), can be stated with the pilot customer's written permission and with the numbers exactly as measured. Guarantees — a refund, a credit for a missed pickup, a first month free — are claims the venture can always substantiate, because it controls them; Book A6 treats guarantees and risk reversal in full. And the stage can be stated plainly: "in development, first sessions in January", which is both lawful and, to an early adopter, attractive.

4.7.2 What you cannot say yet

Four kinds of claim are off-limits until proved. Results the offer has not produced ("save ten hours a month") — until measured on real customers. Comparatives against a named or implied alternative ("faster than any app") — until tested. Numbers of customers, users or successes that do not exist ("trusted by hundreds"). And testimonials that were not given, or were given for something else. Each of these appears constantly on start-up pages, and each is a liability the venture does not need, because Section 4.7.1 offers honest alternatives that early adopters prefer.

4.7.3 The claims file

Every claim in the proposition, on the pre-order page and in the pitch gets a line in a claims file: the claim as worded, the evidence that supports it, the date, and the person who approved it. The file is opened in the discovery quarter, because that is when the first claims are made, and it becomes the compliance record that Book B1's Chapter 10 requires of the marketing system. A claim without a line in the file is not published.

WHO DOES THIS Owning the proposition In a solo venture the founder writes the profile, the map and the statement, and tests them on strangers because there is no one else to disagree; the claims file is a page in the log. In a founding team, the member who owns the customer conversations writes the profile, the member who owns the product writes the map, and the two argue item by item — the argument is the point — before the team drafts the statement together and tests it; one person approves every claim. In an established business, the intrapreneur writes the proposition for the new offer separately from the company's existing brand promise, tests it with customers who do not know the company as well as with those who do, and clears the claims with whoever approves the company's advertising, because the company's name carries the liability. In all three, the proposition is written for one segment, and a second segment gets a second proposition rather than a broader one.

Costly Mistakes

1. Writing the proposition before the profile. The statement describes what the founder wants to build, and the profile is then fitted to it, item by item, until the interviews are forgotten. The consequence is a proposition that passes no repeat test and a product built for an imagined customer.
2. Listing features as value. "Cloud-based, all-in-one, easy to use" asks the customer to translate the product into their own outcome, and most will not. The consequence is a page and a pitch that convert near zero however good the product is.
3. Ignoring the switching cost. A proposition that maximizes the gain and never mentions the fear of change loses the switching test by the 9× mechanism, and the founder concludes that customers are irrational. The consequence is months of "education" spent where a guarantee or a free first month would have sold.
4. One proposition for two segments. The average of two profiles resonates with neither; Norvale's "affordable and compliant" sold nothing for two months. The consequence is a longer sales cycle and a founder who cannot tell which segment is buying.
5. A differentiator nobody values. The "only" sentence is true and the item is not on any customer's top-ranked list. The consequence is a product that is genuinely different and genuinely unsold.
6. Claims without proof. Results not yet produced, comparatives untested, customers who do not exist, on the first pre-order page. The consequence is exposure under the Competition Act and, more immediately, the loss of the trust of early adopters — the only customers a new venture has.

Key Takeaways

• Value is what the customer perceives they receive minus what they perceive they give, relative to the alternative they use today; a benefit the customer does not notice does not exist, and a cost they fear counts even if it is small.
• The customer profile — jobs, pains, gains, ranked — is written from the patterns of the validation log, in the customer's words; a value map — products and services, pain relievers, gain creators — is then written to answer the profile, item by item.
• Fit has three levels, reached in order: problem–solution fit on interview evidence, product–market fit on use and retention, business-model fit on the numbers; a venture that has the first may not claim the second.
• The switching test is arithmetic: perceived gain minus perceived cost, with losses weighted about three times more than gains; the item that wins the test is often a guarantee or a free start, not a feature.
• Differentiation sells when it is a resonating focus — one or two points of difference that matter most to the customer, with proof — and passes four tests: valued, believable, provable, defensible.
• The statement is written in at least two formats and tested with five strangers who must say it back; the format the customer repeats wins.
• Every claim is provable or marked as a hypothesis, kept in a claims file with its evidence and its date, because the Competition Act applies to a founder's first page as it does to a company's advertising.
• Honesty about the stage is a differentiator: research findings stated as research, the founder's credentials, the mechanism, the guarantees and "in development, first delivery in January" are what early adopters trust.

Your Action Plan

The plan below takes about eight hours over two weeks and produces the proposition that Chapters 5, 6 and 9 will use. Steps 1 and 2 fill the two panels from the log; step 3 runs the switching test; steps 4 and 5 write and test the statement; step 6 opens the claims file that every later page will need.

In La Machine

La Machine carries the proposition once it is written, in the places where customers will read it, and it keeps the messaging consistent across them.

• Templates (business/marketing/templates) hold the email templates that every campaign and sequence uses; the proposition's statement, its top pain relievers and its proof lines are written once into the templates and reused, so that the pre-launch messages of Chapter 9 carry the same words as the interviews produced.
• Calls to action (business/marketing/ctas) define the next step a reader is offered — join the waiting list, book a discovery call, pre-order — and each one is written from the proposition for its stage; a call to action that promises more than the claims file allows is caught here before it reaches a page.
• Branding (business/branding) records the brand promise, the tone and the visual charter; for a new venture the promise is the value proposition's outcome in one line, and the page is where it is written so that a designer or a freelancer produces on-message work (→ see Book B3, Branding and Visual Identity).
• Landing pages (business/marketing/landing-pages) will carry the proposition in the tests of Chapter 9; a page is served publicly only when it is published and active, and because publishing is a status rather than a lock, the business's written rule that a person approves each publication is the moment the claims file is checked (→ see Book N5).

What La Machine does not do: it has no customer-profile or value-map canvas — the two panels of Figure 4.1 are drawn on a whiteboard tool or in a document attached to the project — and it does not test statements or verify claims; the repeat test is done with five people, and the claims file is the founder's responsibility. The full treatment of positioning and messaging hierarchies is in Book B1, Chapters 5 and 6; the design of the offer around the proposition — tiers, guarantees, service design — is Book A6.

Self-Check

1. Define value as this chapter uses it, and explain why a benefit the customer does not perceive "does not exist".
2. What are the three lists of the customer profile and the three lists of the value map, and what is the rule for admitting an item to the profile?
3. Name the three levels of fit, the evidence each requires, and why a venture may not claim the second on the evidence of the first.
4. Explain the 9× problem and what it implies for the design of a value map.
5. Give two kinds of claim a venture may lawfully make before it has customers, and two it may not, with the rule that governs them.

Answers. (1) Value is the customer's overall assessment of what they receive minus what they give, as they perceive it and relative to their alternative (Zeithaml, 1988); since the customer's perception decides the exchange, a benefit they do not notice cannot influence it (Section 4.1.1). (2) Profile: jobs, pains, gains, ranked; map: products and services, pain relievers, gain creators; an item enters the profile only if it appeared as a pattern or a strong signal in the validation log, in the customer's words, otherwise it is an italicized hypothesis (Sections 4.2–4.3). (3) Problem–solution fit on interview and observation evidence; product–market fit on use, payment and retention of a real offer; business-model fit on the unit economics; the evidence of the first — people care about the problem and the design addresses it — says nothing about whether they will use, pay for and keep the offer (Section 4.4). (4) Customers weigh losses about three times more than gains, and developers overvalue their product's benefits about three times, a mismatch of about nine (Gourville, 2006); a map must therefore lower the perceived cost and risk of switching — guarantees, a free start, simple onboarding — not only raise the gain (Sections 4.1.3, 4.4). (5) Lawful: research findings stated as research ("in twelve interviews, nine…"), the founder's credentials, the mechanism, guarantees, the stage; unlawful until proved: results not produced, comparatives untested, non-existent customer numbers, testimonials not given — under Competition Act s. 74.01(1)(a) and, for performance claims, s. 74.01(1)(b) (Section 4.7).

Chapter References

Competition Act, R.S.C. 1985, c. C-34, ss. 74.01, 74.1. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-34/

Anderson, J. C., Narus, J. A., & van Rossum, W. (2006). Customer value propositions in business markets. Harvard Business Review, 84(3), 90–99.

Andreessen, M. (2007, June 25). The only thing that matters [Blog post]. Pmarchive. Retrieved September 7, 2026, from https://pmarchive.com/guide_to_startups_part4.html

Competition Bureau. (n.d.). Deceptive marketing practices. Government of Canada. Retrieved September 7, 2026, from https://competition-bureau.canada.ca/

Ellis, S., & Brown, M. (2017). Hacking growth: How today's fastest-growing companies drive breakout success. Crown Business.

Gourville, J. T. (2006). Eager sellers and stony buyers: Understanding the psychology of new-product adoption. Harvard Business Review, 84(6), 98–106.

Moore, G. A. (2014). Crossing the chasm: Marketing and selling disruptive products to mainstream customers (3rd ed.). HarperBusiness.

Osterwalder, A., Pigneur, Y., Bernarda, G., & Smith, A. (2014). Value proposition design: How to create products and services customers want. Wiley.

Zeithaml, V. A. (1988). Consumer perceptions of price, quality, and value: A means-end model and synthesis of evidence. Journal of Marketing, 52(3), 2–22.

Chapter 5 — Market Sizing and Segmentation

"The market is worth eleven billion dollars, and we only need one percent of it." Every lender, investor and experienced founder has heard that sentence, and every one of them has stopped listening at the word only. This chapter replaces it with a number you can defend: how many people have the job, how many of them you can reach, how many you can win in three years — and which fifty of them you will win first.

What You'll Be Able to Do

• Explain what market sizing is for — the decisions it informs and the questions it cannot answer — and avoid the big-number trap.
• Compute TAM, SAM and SOM bottom-up, from counts, prices and frequencies you can source, and state each with its assumptions.
• Segment a market for a venture that has no customers yet, using the criteria that matter for a new offer and a segment grid that ranks them.
• Write a first ideal customer profile — fit, value, reachability, exclusions — consistent with the definition used across this library.
• Choose a beachhead: the first segment to win, selected on eight criteria, from which the next segments follow.
• Source the numbers from public statistics, associations and directories, and triangulate them honestly.
• Turn the obtainable market into a first-year target and a weekly number that the discovery sprint and, later, the sales process can act on.

Why It Matters

Market sizing answers the question that precedes every other investment decision: is this worth doing, and at what scale? A venture that sizes honestly avoids two opposite failures — the market too small to sustain the effort, discovered after a year of work, and the market so large that the founder cannot say where to start. Segmentation answers the second question by choosing where to start, and the beachhead turns that choice into a plan a solo founder can execute in a year. For a team, sizing is the argument that ends the debate about ambition; for an established business, it is the case that gets a new offer a budget. And for anyone who will ask a lender or an investor for money, it is the page they read first — and the page on which "one percent of eleven billion" ends the conversation.

In This Chapter

• 5.1 Why Size the Market
• 5.2 TAM, SAM and SOM
• 5.3 Segmentation for a New Venture
• 5.4 The Ideal Customer Profile Before You Have Customers
• 5.5 The Beachhead Market
• 5.6 Sourcing the Numbers
• 5.7 From Sizing to Targets

5.1 Why Size the Market

A founder sizes a market to make four decisions, not to impress anyone. Is it worth it? A problem that is frequent and intense for two hundred people in the region may be a good freelance practice and a poor company; the size decides which one the founder is building. Which segment first? Sizing several candidate segments side by side is what makes the beachhead choice of Section 5.5 possible. What is the target? The obtainable market, divided by the years it takes to win it, is the first revenue target and the first capacity plan. And how much to invest? A venture cannot rationally spend more to build and reach a market than the market can return, and the size sets that ceiling before the plan of Book A2 sets the budget.

5.1.1 Two ways to size, and the one that works

Sizes are computed top-down or bottom-up. Top-down starts from a large published number — an industry's total revenue, a category's global sales — and applies a share the founder hopes to capture. It is fast, it is what the "one percent" sentence does, and it is nearly worthless for a new venture, because the share is the whole question and the published number usually describes a different market from the one the venture serves. Bottom-up starts from the smallest unit — one customer, one purchase — and multiplies up: how many people have the job, what share of them fit the offer, what each will pay, how often. It is slower, it requires sources, and every number in it is a hypothesis that Chapter 3's evidence can test. Steve Blank and Bob Dorf, and Bill Aulet after them, made bottom-up sizing a standard step of disciplined venture building for exactly that reason: it produces numbers a founder can defend and revise (Blank & Dorf, 2012; Aulet, 2013).

KEY CONCEPT Bottom-up beats top-down, because every number in it can be tested A top-down size is a large number multiplied by a wish. A bottom-up size is a chain of small numbers — a count of people with the job, a share who fit, a price, a frequency — each of which comes from a source and each of which the interviews, the surveys and the tests of this book can confirm or correct. The bottom-up size is smaller, and it is the only one that changes what the founder does on Monday: it says how many people must be reached, at what rate, to hit a target. A founder who can show the chain, with its sources and its assumptions marked, is believed; one who shows a share of a billion is not.

5.1.2 The big-number trap

The trap works in three moves. A published total makes the market look enormous; a small share makes the ambition look modest; and the product of the two is a revenue figure with no path behind it. Lenders and investors know the move and discount it to zero; the more damaging effect is on the founder, who plans production, hiring and spending for a number that no channel can deliver. The antidote is the obtainable market of Section 5.2, which is bounded not by ambition but by the channels the venture actually has and the capacity it can actually build.

WATCH OUT "We only need one percent" The sentence hides four errors at once. The published market is usually a different market — all bookkeeping in Canada is not the market for a quarterly receipt clinic in one region. One percent of anything is not a plan, because no channel produces "one percent"; channels produce contacts at a rate, and contacts convert at a rate. The one percent is often larger than the venture's capacity to deliver, which nobody checks. And the sentence signals to every experienced reader that the founder has not spoken to customers, because a founder who has speaks in counts of people. Replace it with a bottom-up chain and a beachhead, and the same modesty becomes credible.

5.2 TAM, SAM and SOM

Three nested numbers organize a bottom-up size, and Figure 5.1 draws them for the receipt-clinic case. They are defined here once, because the terms are used loosely in practice and the loose usage is a source of the trap above.

DEFINITION TAM, SAM and SOM The total addressable market (TAM) is the annual value of the job across everyone who has it, if every one of them bought a solution like yours at your price — the size of the problem, not of your business. The serviceable addressable market (SAM) is the part of the TAM that your offer, as designed, and your channels, as they exist, can actually reach: the region you can serve, the segments the offer fits, the people your channels touch. The serviceable obtainable market (SOM) is the part of the SAM you can realistically win in a stated period — usually three years — given your capacity to deliver, your channels' throughput and the competition; for a service business it is usually bounded by capacity, for a product business by channel reach and conversion. All three are annual values, stated with their assumptions, and the SOM is the only one that appears in a first-year plan.

Figure 5.1 shows the three sizes as nested areas rather than as a funnel, together with the beachhead that sits inside the obtainable market. It is read from the outside in: the total addressable market as the outer area, the serviceable addressable market inside it, the serviceable obtainable market inside that, and the beachhead of Section 5.6 as a small, named region within the obtainable market. The nesting is the point — each figure is a subset of the one before it, computed bottom-up from the assumptions listed beside it, never a percentage taken of the one above.

Figure 5.1 — TAM, SAM, SOM and the beachhead. The total addressable market is everyone with the job; the serviceable addressable market is what the offer and the channels can reach; the serviceable obtainable market is what can be won in three years, here bounded by delivery capacity; the beachhead is the first segment won inside it. Original figure; illustrative figures from the fictional receipt-clinic case.

5.2.1 The bottom-up chain

Each of the three numbers is a chain of counts and rates. For the TAM: the number of people or businesses with the job, from a public count; the share who would use a solution of this kind, from interviews or a survey; the annual value per customer, from the price hypothesis of Chapter 7 times the frequency of purchase. For the SAM: the same chain restricted to the geography the offer can serve, the segments it fits and the channels that exist. For the SOM: the SAM times the share the venture can win in three years — or, more honestly for most first ventures, the number of customers the venture can deliver to, times the annual value. Every link is written down with its source and marked sourced, estimated or guessed, exactly as the scorecard of Chapter 2 marked its scores.

DO THE MATH TAM, SAM and SOM for a regional service Illustrative figures; the counts are from public statistics as of 2025, the shares and prices are hypotheses to be tested. The receipt clinic of Chapters 2 to 4 sells four two-hour sessions a year at $180 CAD each, or $720 per client per year. TAM. Canada counted 1,079,188 small employer businesses (1 to 99 employees) as of December 2024 (Innovation, Science and Economic Development Canada, 2025). Suppose that about 45% of them use an external bookkeeper or accountant for their books — a share to be verified by the survey of Chapter 3 — giving about 486,000 businesses with the job. TAM ≈ 486,000 × $720 ≈ $350 million a year: the size of the problem, not of Mariam's business. SAM. Mariam serves one metropolitan region in person or by video. The Canadian Business Counts give the number of employer businesses in that region — say 28,000 — and their distribution by size; the interviews placed the segment at owners above $300,000 of revenue with staff and no in-house assistant, which the counts approximate as businesses with 5 to 49 employees, about 35% of the region's employer businesses, or 9,800. At the same 45% external-bookkeeping share, the SAM is about 4,400 businesses, or 4,400 × $720 ≈ $3.2 million a year. SOM. A service is bounded by capacity: one bookkeeper can run about 40 clinic clients a year (160 sessions of two hours, plus preparation, alongside ordinary work). With a second bookkeeper hired in year two, the practice can serve about 80 clinic clients by year three: SOM ≈ 80 × $720 ≈ $57,600 a year, or 1.8% of the SAM. That share is not an ambition; it is what two people can deliver. It tells Mariam that the clinic is a profitable line for her practice and a poor basis for a company — unless the model changes (a productized kit, a license to other bookkeepers), which Chapter 6 explores. The chain, written down with its sources, is the whole argument.

5.2.2 Capacity-bounded and channel-bounded obtainable markets

For most service ventures, the obtainable market is bounded by what the founder and the first hires can deliver, as in the example, and the sizing exercise reveals early whether the founder is building a practice or a company. For product ventures — software, a physical product, a course — capacity is less binding and the SOM is bounded by the channels: how many of the SAM the venture's channels can reach in three years, times the conversion rate those channels achieve. A software version of the clinic — a guided quarterly sorting kit sold to bookkeepers — would have the same TAM, a national SAM limited by the channels that reach bookkeepers (an association of 1,200 members, a trade publication, search), and a SOM of perhaps 1,200 × 8% conversion over three years ≈ 96 customers, each of whom serves 40 clients. The two SOMs describe two different businesses from one opportunity; Chapter 6's patterns make the choice explicit.

5.2.3 Stating the size honestly

A size is stated as a range with its assumptions, never as a point. "SAM between $2.6 million and $3.9 million a year, depending on the share of businesses using external bookkeeping (35% to 55%, to be measured in the survey)" is a sentence a lender can work with; "$3.2 million market" is not. The assumptions table of Chapter 7 (Table 7.2) records every link of the chain with its range, and the volatile-facts register of the bibliography records the public counts with their dates, because business counts are revised twice a year.

5.3 Segmentation for a New Venture

A segment is a group of customers who share the job, the pains, the buying behavior and the channels closely enough that one proposition, one price and one path to reach them serve all of them. Book B1 (Chapter 3) developed segmentation for a business that has customers; this section adapts it to a venture that has interviews.

5.3.1 Criteria that matter for a new offer

Marketing textbooks list the classic criteria — demographic, geographic, firmographic, behavioral, psychographic (Kotler & Keller, 2016) — and Thomas Bonoma and Benson Shapiro organized the business-market criteria into a nested set, from the outer, easily observed layers (industry, size, location) to the inner, decisive ones (purchasing approach, situational factors, the buyer's personal characteristics) (Bonoma & Shapiro, 1983). For a new venture, four criteria do most of the work, and they are ordered from the inside out. Needs-based: the ranking of jobs, pains and gains in the profile — the segment split of Chapter 3 was a needs-based split, discovered at about $300,000 of revenue. Situational: the trigger event that makes the job urgent — a tax deadline, a move, a succession, an audit, a new regulation — because a segment defined by its trigger can be reached at the moment it is ready to buy. Behavioural: the workaround in use, the tool already adopted, the professional already hired — because behavior predicts willingness to switch. And only then firmographic or demographic: industry, size, revenue, location, age, household — the observable proxies that make the segment countable in public statistics and findable in a directory.

The order matters because the observable criteria are only useful as proxies for the decisive ones. "Businesses with 5 to 49 employees" is not a segment; it is the countable approximation of "owners who feel March as a crisis and have no in-house help", which is the segment. A founder who segments from the outside in ends up with a group that is easy to count and does not share the pain.

5.3.2 The segment grid

Table 5.1 is the grid on which candidate segments are compared. Each row is a criterion that predicts whether the venture can win a segment — size (the count from Section 5.6), pain intensity (from the log), willingness to pay (from commitments), reachability (channels that exist), the alternatives in place (from Chapter 8's grid), and the founder's access — and each of the three columns that follow is one of the segments Mariam's interviews revealed, carrying its evidence and its score from 1 to 5. The last row totals the scores. Read the grid as a comparison, not as a verdict: the totals rank the segments, and the beachhead criteria of Section 5.5 then decide which one comes first.

Table 5.1 — The segment grid (example rows from the receipt-clinic case, scores 1 to 5)

5.3.3 Business and consumer segments

Business segments are counted in firmographic units — industry, size, location — and reached through the channels businesses use: associations, suppliers, professionals, trade events, search. Their decision units have several members (Chapter 3), and a segment is sometimes defined by who decides rather than by the business — the kitchen manager and the owner of Norvale's case were two segments inside one restaurant. Consumer segments are counted in demographic units — age, household, income, location — and reached through the channels people use; their triggers are life events (a move, a birth, a retirement, a diagnosis) and their decision units are households. In both cases the needs-based split comes from the interviews, and the observable criteria are chosen afterwards to count and find the people who share it.

5.4 The Ideal Customer Profile Before You Have Customers

Book B1 defined the ideal customer profile — the ICP — for a business that has customers, as a one-page description of the customers the business should pursue, on four dimensions: fit (they have the problem the offer solves, in the circumstance it solves it), value (they are worth serving — revenue, margin, retention, referrals — relative to the cost of serving them), reachability (the business can find and reach them through channels it has), and exclusions (who the business will not pursue, written down). A venture without customers uses the same four dimensions and the same one-page format, so that the ICP written in discovery becomes, without translation, the ICP that Book B1's marketing system and Book E1's sales process will use. Three things change in the sources.

Fit is written from the profile of Chapter 4. The ICP's fit criteria are the observable proxies for the top-ranked jobs and pains: for the clinic, "revenue above $300,000, staff on payroll, no in-house bookkeeping, files with an external accountant" stands in for "feels March as a crisis and has nobody to delegate it to." The proxies are named as proxies, so that a prospect who fits the proxy but not the pain is recognized in the first conversation.

Value is estimated, not measured. Without a customer history, value is the annual value from the sizing chain, the expected retention (a quarterly service with a routine is retained; a one-off sorting is not), the referral potential (the association) and the cost to serve (two hours of a bookkeeper's time per session, plus travel). It is written as a hypothesis with a range, and the first ten customers will correct it.

Exclusions are the most important line. A venture without customers is tempted to serve anyone who asks, and the first year is spent on the wrong customers if the exclusions are not written before the first sale. Mariam's exclusions read: owners under $150,000 (the pain is too mild for the price), businesses with in-house bookkeeping (the job is done), businesses outside the region (the model is in-person or by video with a local accountant relationship), and — from the interviews — "anyone who wants an app", because the clinic is a service and the app-seeker will churn in June.

DEFINITION Ideal customer profile (for a new venture) The ideal customer profile is the one-page description of the customers the venture will pursue first, on four dimensions — fit, value, reachability and exclusions — written from the evidence of discovery and consistent with the definition used across this library (Book B1, Chapter 3). Fit names the observable proxies for the top-ranked pains; value estimates annual value, retention, referral potential and cost to serve, with ranges; reachability names the channels that exist today; exclusions name who will not be pursued, and why. The ICP is signed by the founder, dated, and revised after the first ten customers.

CASE IN POINT A consultant choosing between three segments Fictional example; names and figures are invented. Daniel Okafor, a human-resources consultant with twelve years in a large company, left to start a practice helping small businesses hire their first employees properly — contracts, onboarding, the rules of the Act respecting labour standards. His twelve interviews split into three segments. Dental and medical clinics hiring receptionists and hygienists: high pain (a bad hire cost one clinic $18,000 and a lawsuit threat), a budget, a professional association with a newsletter, but a decision unit that included an office manager who felt threatened by an outside consultant. Construction contractors hiring their first crew: very high pain and urgency (a job lost for lack of hands), reachable through a suppliers' counter and a regional association, but a strong preference for "a guy I know" over a consultant and a low willingness to pay for paper. Non-profits hiring a coordinator: high need, gratitude, reachable through a network of funders, and almost no budget. The segment grid scored the clinics highest on willingness to pay and reachability, the contractors highest on pain and urgency, the non-profits highest on access. Daniel chose the clinics as his beachhead — the budget and the association decided it — wrote the contractors as the second segment with a different proposition ("a crew hired in two weeks, papers done"), and served non-profits at a reduced rate as a source of referrals and testimonials, explicitly, in the exclusions section: "non-profits: two per year, at cost, for reputation; not a growth segment." The ICP was one page. It was the most useful page he wrote that year.

5.5 The Beachhead Market

Geoffrey Moore used the image of a beachhead — the first strip of shore an invading force takes and holds before moving inland — for the first segment a venture must win completely before it can win any other, and he described the sequence of adjacent segments that follows as a bowling alley: each pin knocked down tips the next (Moore, 2014). Bill Aulet made the beachhead the central decision of his method for new ventures, with a set of criteria for choosing it (Aulet, 2013). The idea is the practical answer to the founder's fear of being too small: a venture that owns a segment of two hundred customers who talk to each other has references, revenue and a model that works, and can expand; a venture that has ten customers in each of ten segments has none of those things.

DEFINITION Beachhead market The beachhead market is the first segment the venture chooses to win, and to win decisively, before any other: a group small enough to dominate with the resources at hand, coherent enough that its members talk to each other and buy the same whole product, valuable enough to sustain the venture while it learns, and positioned so that winning it opens adjacent segments. The beachhead is a subset of the obtainable market of Section 5.2, chosen on the eight criteria of Table 5.2; it is where the ideal customer profile, the value proposition and the first-year target all point.

5.5.1 The eight criteria

Table 5.2 lists the criteria on which candidate beachheads are compared, adapted from Aulet's method and from the experience of small ventures that are not software start-ups. Each criterion is scored from 1 to 5, and the table states for each one what the score means and how a founder finds the evidence. Read the criteria as filters as much as scores: a segment that scores 1 on "the customer has the money" or on "we can deliver the whole product today" is disqualified whatever its total, because a beachhead that cannot pay or cannot be served is a beach the venture cannot land on.

Table 5.2 — Beachhead selection criteria (scored 1 to 5; a 1 on criteria 1, 3 or 4 disqualifies)

METHOD The beachhead selection grid (two hours) When to use it: once the segment grid of Table 5.1 has ranked two to four candidate segments, before the canvas of Chapter 6 is drawn. Steps: (1) Copy the two to four highest-ranked segments from Table 5.1 as columns. (2) Score each on the eight criteria of Table 5.2, from evidence in the log; mark guessed scores. (3) Disqualify any segment with a 1 on criteria 1, 3 or 4, whatever its total. (4) For the remaining segments, write in one sentence the adjacent segment each one would open (criterion 7) — a beachhead with no next pin is a niche, which may be fine for a practice and is not for a company. (5) Choose the highest total among the qualified; if two are close, choose the one where members talk to each other most, because word of mouth is the cheapest channel a new venture will ever have. (6) Write the choice, the runner-up and the reasoning in the decision registry, dated. Deliverable: the scored grid, the choice, the runner-up, the adjacent segment named. Pitfalls: choosing the largest segment (size is not a criterion here — the segment grid already counted it); choosing the segment the founder finds most interesting rather than the one with money and urgency; scoring "we can deliver the whole product" from ambition rather than from today's capacity; forgetting to write the runner-up, which is the pivot of Chapter 10 if the beachhead fails.

5.5.2 Owning the beachhead

A beachhead is won when the venture is the obvious choice for that segment — when a member of the association who mentions the problem is told "call Mariam" by another member. Practitioners describe the signal as a dominant share of the segment's new purchases, typically more than half, and a referral rate that makes outbound effort optional. For a small venture that means concentrating every hour of marketing and sales on the beachhead for a year, refusing adjacent customers politely or serving them at the margin, and measuring share in the segment rather than revenue in general. Book B1's marketing system and Book E1's sales process are built to be pointed at one segment; the beachhead is what they are pointed at.

5.5.3 The bowling alley

The adjacent segments are chosen on criterion 7 and written into the ICP as "next": the segment that shares the pain but uses a different channel, or shares the channel but has a related pain. Mariam's next pin was the accountants who referred her clients — the same relationship, a different buyer, a licensed kit rather than a service; Daniel's was the contractors, the same whole product with a different proposition. Naming the next pin in the discovery quarter does not commit the venture to it; it commits the venture to a direction, which is what a lender means by "scalable" and what a co-founder means by "where is this going?".

5.6 Sourcing the Numbers

A bottom-up chain is only as good as its counts, and the counts a founder needs are, for the most part, public and free. This section names them and explains how to combine them.

5.6.1 Public statistics

Business counts by industry, size and region are the backbone of any business-market size, and population counts by age, household and region are the backbone of any consumer-market size. In Canada, both come primarily from Statistics Canada, with provincial detail from the Institut de la statistique du Québec (2026); the Québec and Canada Note below gives the specifics. The counts are classified by the North American Industry Classification System, and a founder's first task is to find the NAICS codes that approximate the segment — a restaurant is 722511 (full-service) or 722512 (limited-service), a bookkeeping practice is 541215 — because every count is organized by those codes (Statistics Canada, 2022b).

5.6.2 Associations, directories and the venture's own count

Public statistics give the count of businesses in a code, not the count with the pain. Three sources narrow it. Associations publish or will state their member counts, which are both a size and a channel: an association of 1,200 members is a SAM component and a reach estimate at once. Directories — the business directories and map sources that La Machine's Business Explorer searches, the professional registers, the sector lists — give the venture a countable list of names, which is the strongest form of a size because each entry can be contacted. And the venture's own count — the interviews and the survey — gives the share of the count that has the pain, uses the workaround and would pay; that share, applied to the public count, is the SAM.

5.6.3 Triangulation

Every size in this chapter is computed twice, from two independent chains, and the difference is reported. A regional SAM computed from the Business Counts times the survey share, and the same SAM computed from the association's membership times the referral rate of the interviews, will not agree; if they are within a factor of two, the range is the size, and if they are not, one chain has a broken link that the founder must find before a lender does. Triangulation is also the check on the shares that no source provides — "45% use an external bookkeeper" — which should come from at least one survey question and one external indicator (a professional body's client count, a tool vendor's published user base, an industry study).

QUÉBEC & CANADA NOTE Counting businesses and people in Canada (as of September 2026) Statistics Canada's Canadian Business Counts (Table 33-10-0764-01 for businesses with employees, and a companion table for businesses without) give the number of businesses by NAICS code, by employee-size class and by geography, down to the census subdivision and the census metropolitan area; they are drawn from the Business Register and released twice a year — the December 2024 counts are the reference as of this writing, and a new release should be checked before quoting (Statistics Canada, 2025). Census profiles give population by age, household type, income, language and occupation for every municipality and neighborhood, and the 2021 Census is the current reference until the 2026 Census results are released (Statistics Canada, 2022a). ISED's Key Small Business Statistics summarize the national picture — 1,079,188 small employer businesses as of December 2024 — and the survival curve used in Chapter 1 (Innovation, Science and Economic Development Canada, 2025). The Institut de la statistique du Québec publishes Québec's regional economic and demographic profiles, employment by industry and region, and business demography. The Registraire des entreprises holds the public register of Québec enterprises, searchable by name and by activity code, which the venture may consult for individual businesses (→ see Book A4). And NAICS Canada 2022 is the classification that organizes every business count; its codes are the first thing to note in the sizing sheet (Statistics Canada, 2022b). All of these are free; the Where to Get Help section of the master bibliography lists their addresses.

5.7 From Sizing to Targets

The obtainable market is a three-year number. The founder needs a first-year number, and then a weekly one, because a venture is run by the week.

5.7.1 The first-year target

The first-year target is the share of the SOM that the venture's capacity and channels can deliver in twelve months, and for most ventures it is bounded by the same constraint that bounded the SOM — capacity for a service, channel throughput for a product — with a ramp: the first quarter is discovery, the second builds the beachhead's first customers, the third and fourth run at rate. Mariam's SOM of 80 clients by year three becomes a first-year target of about 30, because the first quarter produces the model and the pilot, the second produces the first ten clients from the deposits and referrals, and the last two quarters add ten each. Daniel's becomes fifteen clinics, because each engagement takes three weeks and he is alone.

5.7.2 The weekly number

A first-year target becomes a weekly number by walking the funnel backwards, using the arithmetic that Book B1 (Chapter 2) applies to a marketing system and that a venture in discovery estimates from its interviews. The Do the Math box shows the walk; the result — a number of people to contact each week — is what the discovery sprint of Chapter 11 and, later, the sales process of Book E1 are built to produce. A weekly number that exceeds the hours available is the earliest possible warning that the target, the channel or the model must change.

DO THE MATH From the obtainable market to Monday morning Illustrative figures. Mariam's first-year target is 30 clinic clients. From the interviews and the first deposits, she estimates the rates: of the owners who fit the ICP and are contacted, about 25% agree to a conversation; of those, about 50% book a first session after the conversation; and of those, about 80% continue after the first session — the first-year "client" being one who completes at least two sessions. Working backwards: 30 clients ÷ 0.80 = 38 first sessions; 38 ÷ 0.50 = 75 conversations; 75 ÷ 0.25 = 300 fitting owners contacted over the year. Removing the first quarter of discovery, that is 300 over 39 weeks, or about 8 contacts a week — 2 hours of a 6-hour slot at 15 minutes each, which is feasible. Now the product version: 96 bookkeeper-customers in three years, 30 in year one, through an association newsletter that reaches 1,200 members with a typical 3% response and a 25% conversion from response to purchase: 1,200 × 0.03 × 0.25 = 9 customers per campaign — so four campaigns a year and a second channel are needed for 30, which the calendar of Chapter 11 must hold. Either way, the weekly number is the test of the target: if 8 contacts a week cannot be found in the channels that exist, the SOM was a wish.

WHO DOES THIS Sizing and choosing in the three situations In a solo venture the founder computes the chains, and the discipline is to write every assumption down and to check the chain with one outsider who has run a business in the segment, because the founder's estimates of shares and conversion rates are systematically optimistic. In a founding team, one member owns the public counts and the sources, another the shares from the interviews, and the two chains are triangulated before the team scores the beachhead — the disagreement between the chains is the most useful conversation of the week. In an established business, sizing a new offer starts from the company's own customer base — how many current customers fit the new profile — before any public count, because the existing base is both the most reliable count and the cheapest channel; the intrapreneur then sizes the market beyond it, and presents both. In all three, the beachhead decision is recorded with its runner-up, because the day it fails, the runner-up is the plan.

Costly Mistakes

1. Sizing top-down and calling a share a plan. "One percent of eleven billion" ends the conversation with any experienced reader and, worse, sets production and spending for a number no channel can deliver. The consequence is a lost year of capacity built for customers who never arrive.
2. Confusing the size of the problem with the size of the business. The TAM measures how many people have the job; the SOM measures what the venture can win; presenting the first as the second misleads the founder before it misleads anyone else. The consequence is a service practice that believed it was a company, or a company that starved itself as if it were a practice.
3. Segmenting from the outside in. "Businesses with 5 to 49 employees" is a proxy, not a segment; a founder who treats it as the segment sells to people who share the size and not the pain. The consequence is a conversion rate that stays low for reasons the numbers cannot explain.
4. Serving everyone who asks. Without written exclusions, the first year goes to the wrong customers — the app-seekers who churn in June, the non-profits without a budget — and the beachhead is never owned. The consequence is revenue without references and effort without compounding.
5. Choosing the beachhead for its size or its interest. Size is already counted in the segment grid; the beachhead criteria are money, urgency, deliverability, channel and word of mouth. The consequence of choosing the largest or the most interesting segment is a beach the venture cannot land on.
6. A target without a weekly number. A first-year target that was never walked back to contacts per week is a hope; the first warning that it exceeds the hours available arrives in month nine. The consequence is a founder who worked hard on the wrong volume for three quarters.

Key Takeaways

• Market sizing informs four decisions — whether the venture is worth it, which segment first, what the target is and how much to invest — and it is done bottom-up, because every link of a bottom-up chain can be sourced and tested.
• TAM is the size of the problem, SAM is what the offer and the channels can reach, SOM is what can be won in three years — bounded by capacity for a service and by channel throughput for a product — and only the SOM belongs in a first-year plan.
• A size is stated as a range with its assumptions, each link marked sourced, estimated or guessed, and computed twice from independent chains.
• Segments are defined from the inside out — needs, trigger, behavior — and the observable criteria are chosen afterwards as countable proxies; a proxy that does not track the pain is not a segment.
• The ideal customer profile of a new venture has the same four dimensions as the library's — fit, value, reachability, exclusions — written from discovery evidence, and the exclusions are its most important line.
• The beachhead is the first segment to win decisively, chosen on eight criteria — money, reach, urgency, deliverability, competition, word of mouth, adjacency, fit with the founder — with a runner-up recorded as the pivot.
• The numbers come from public statistics organized by NAICS code, from associations and directories, and from the venture's own survey share; they are free, dated and revised twice a year.
• The obtainable market becomes a first-year target and a weekly number by walking the funnel backwards; a weekly number that exceeds the hours available is the earliest warning that the target must change.

Your Action Plan

The plan below takes about ten hours over two weeks and produces the size, the segment choice, the ICP and the weekly number that Chapters 6, 7 and 11 use. Steps 1 and 2 build and check the chains; step 3 ranks the segments; step 4 chooses the beachhead; step 5 writes the ICP; step 6 turns the size into a number to act on.

In La Machine

La Machine gives the founder two of the counts and the place to keep the segments, and it computes the arithmetic of the chains.

• The Business Explorer (business/marketing/business-explorer) searches five sources — two map sources and three directories — by name, industry and city, and gathers the results in a staging table. Each search returns at most twenty results on the map sources and two hundred on the directories, which work only when their database is installed, so the count of results for a type in a region is a directory-based size that triangulates the public statistics only while it stays below that cap, and a large region is searched in narrower parts. The list itself is a reachable segment: each entry can be converted into a company prospect under Companies, and a contact created from an email found there arrives with every channel blocked.
• Segments (business/marketing/segments) hold the ICP's fit criteria as a segment definition over the contacts and companies the venture records; the page shows a live count of the contacts that match, which is the venture's own count growing as discovery adds people; the do-not-reach list holds email addresses, domains and telephone numbers with a reason, such as an unsubscribe or a do-not-call request, so it protects the people who asked not to be contacted rather than the groups the ICP excludes.
• Companies (business/marketing/companies) record the businesses of the beachhead with their industry, size and source, so that the weekly number of Section 5.7 is worked from a list rather than from memory.
• Calculator & Tools (productivity/tools) provides the calculators — percentages, margins, break-even and the like — that the chains and the funnel walk-back use; a spreadsheet attached to the project holds the sizing sheet itself.

What La Machine does not do: it holds no public statistics, census data or NAICS tables — those come from the sources of the Québec and Canada Note — and it does not compute TAM, SAM or SOM or score segments; the sizing sheet and the two grids are the founder's documents, attached to the project. Book B1 (Chapter 3) develops the ICP for a business with customers, Book B2 (Market Research and Competitive Intelligence) develops the sources and methods of market research, Book N4 shows how contacts and companies fit together in the platform, and Book N5 (Marketing Automation in La Machine) shows how segments are drawn from them.

Self-Check

1. Why does this chapter insist on bottom-up sizing, and what is wrong with "one percent of an eleven-billion-dollar market"?
2. Define TAM, SAM and SOM, and say which one belongs in a first-year plan and what bounds it for a service and for a product.
3. In what order should a new venture apply segmentation criteria, and why is "businesses with 5 to 49 employees" not a segment?
4. Name the four dimensions of the ideal customer profile and explain why the exclusions are described as the most important line.
5. Which three beachhead criteria disqualify a segment on a score of 1, and why is the runner-up recorded?

Answers. (1) Every link of a bottom-up chain — a count, a share, a price, a frequency — can be sourced and tested by the evidence of this book, whereas a top-down share is the whole question multiplied by a wish; the "one percent" sentence uses a different market, treats a share as a plan, ignores capacity and signals that the founder has not spoken to customers (Section 5.1). (2) TAM is the annual value of the job across everyone who has it; SAM is the part the offer and the channels can reach; SOM is the part that can be won in a stated period; only the SOM belongs in a first-year plan, bounded by delivery capacity for a service and by channel reach and conversion for a product (Section 5.2). (3) From the inside out — needs, trigger, behavior — then observable proxies; a size class is a countable approximation of a segment defined by a shared pain, and people who share the size do not necessarily share the pain (Section 5.3). (4) Fit, value, reachability and exclusions; without written exclusions a venture without customers serves everyone who asks, spends its first year on the wrong customers and never owns its beachhead (Section 5.4). (5) The customer has the money, there is a compelling reason to buy now, and the venture can deliver the whole product today; the runner-up is recorded because it is the pivot of Chapter 10 if the beachhead fails (Section 5.5).

Chapter References

Aulet, B. (2013). Disciplined entrepreneurship: 24 steps to a successful startup. Wiley.

Blank, S., & Dorf, B. (2012). The startup owner's manual: The step-by-step guide for building a great company. K&S Ranch.

Bonoma, T. V., & Shapiro, B. P. (1983). Segmenting the industrial market. Lexington Books.

Innovation, Science and Economic Development Canada. (2025). Key small business statistics 2025. Government of Canada. Retrieved September 7, 2026, from https://ised-isde.canada.ca/site/sme-research-statistics/en/key-small-business-statistics/key-small-business-statistics-2025

Institut de la statistique du Québec. (2026). Statistiques et publications. Gouvernement du Québec. Retrieved September 7, 2026, from https://statistique.quebec.ca/

Kotler, P., & Keller, K. L. (2016). Marketing management (15th ed.). Pearson.

Moore, G. A. (2014). Crossing the chasm: Marketing and selling disruptive products to mainstream customers (3rd ed.). HarperBusiness.

Statistics Canada. (2022a). Census Profile, 2021 Census of Population. Retrieved September 7, 2026, from https://www12.statcan.gc.ca/census-recensement/2021/dp-pd/prof/index.cfm

Statistics Canada. (2022b). North American Industry Classification System (NAICS) Canada 2022 Version 1.0. Retrieved September 7, 2026, from https://www.statcan.gc.ca/en/subjects/standard/naics/2022/v1/index

Statistics Canada. (2025). Canadian Business Counts, with employees, December 2024 (Table 33-10-0764-01). Retrieved September 7, 2026, from https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3310076401

Chapter 6 — The Business Model Canvas

A founder has a validated problem, a tested value proposition, a beachhead and a first ideal customer profile — four documents that describe the customer side of a business. What they do not yet describe is the business: how the offer will reach those customers, what it will take to deliver it, who must help, what it will cost and how the money will come in. This chapter puts all of it on one page, so that the whole model can be read, challenged and tested at once.

What You'll Be Able to Do

• Draw a business model on the nine blocks of the canvas, in the order a founder fills them, and read it as a set of hypotheses rather than as a poster.
• Fill the right side — customer segments, value propositions, channels, customer relationships, revenue streams — from the evidence of Chapters 3 to 5.
• Fill the left side — key resources, key activities, key partners, cost structure — honestly, and recognize when it describes a business the founder cannot run.
• Recognize and choose a business model pattern — subscription, platform, freemium, razor-and-blade, unbundling, long tail, productized service, licensing — and know what each one requires to be true.
• Run a two-hour canvas workshop, and use the Lean Canvas variant when the venture is at the stage it was designed for.
• Read a canvas critically with coherence tests, and find its dependencies and single points of failure.
• Tell the model as a two-minute story that a customer, a co-founder or a lender can repeat.

Why It Matters

The canvas is where the venture becomes one thing instead of several. A founder who holds the customer in one document, the offer in another and the costs in a spreadsheet cannot see that the channel eats the margin, that the key partner is also the main competitor, or that the revenue stream comes from a segment nobody has interviewed. On one page, those contradictions are visible in minutes. For the founder, the canvas is the instrument that turns the discovery quarter's evidence into a business; for a team, it is the shared object that ends debates about what the business is; for an established business launching a new line, it is the page that shows where the new model borrows from the old one and where it must not.

In This Chapter

• 6.1 The Canvas and Its Nine Blocks
• 6.2 The Right Side
• 6.3 The Left Side
• 6.4 Business Model Patterns
• 6.5 Filling the Canvas
• 6.6 Reading a Canvas Critically
• 6.7 From Canvas to Story

6.1 The Canvas and Its Nine Blocks

The Business Model Canvas is a one-page template with nine blocks that together describe how a business creates, delivers and captures value. Alexander Osterwalder developed the underlying framework in his doctoral research on business model ontology and, with Yves Pigneur and several hundred contributing practitioners, published it as the canvas in 2010; it has since become the most widely used way of describing a business model in a single view (Osterwalder & Pigneur, 2010). The template itself is published by Strategyzer AG under a Creative Commons Attribution-ShareAlike license, which is why it appears, with attribution, in so many tools and courses; this book describes the concept and draws its own figure rather than reproducing the template.

DEFINITION Business Model Canvas The Business Model Canvas is a one-page description of a business model in nine blocks — customer segments, value propositions, channels, customer relationships, revenue streams, key resources, key activities, key partners and cost structure — each holding the choices the venture has made or the hypotheses it holds about that part of the model (Osterwalder & Pigneur, 2010). Its purpose is to make the whole model visible at once, so that its parts can be checked against each other and its hypotheses can be tested one by one. It is a working document, versioned and dated, not a finished description.

6.1.1 The nine blocks in reading order

The canvas can be read in any order, but a founder fills it in one particular order, because each block depends on the ones before. Figure 6.1 arranges the nine blocks in that order, in four rows. The first row is who and what: the customer segments the venture serves, and the value propositions it offers each of them — the work of Chapters 3 to 5. The second row is how it reaches them: the channels through which each segment learns about, evaluates, buys, receives and gets support for the offer, and the kind of relationship each segment expects. The third row is what it takes: the key resources the venture must own or control, the key activities it must perform well, and the key partners who supply what it cannot or should not do itself. The fourth row is the money: the revenue streams that come from each segment, and the cost structure that the resources, activities and partners impose. Chapter 7 develops the fourth row into numbers.

Figure 6.1 — The nine blocks in reading order. Who and what: customer segments and value propositions. How it reaches them: channels and customer relationships. What it takes: key resources, key activities and key partners. The money: revenue streams and cost structure. Original figure, adapted from the concept described by Osterwalder and Pigneur (2010); the arrangement follows this book's reading order rather than the canvas template.

6.1.2 A hypothesis board, not a poster

The canvas is often filled once, framed and forgotten, and in that use it is worthless. Its value is as a hypothesis board: every item on it is either a fact the venture has established or a hypothesis it holds, and the two are marked differently — in practice, with two colors of sticky note, or with italics on a page. A canvas filled for the first time in the discovery quarter is almost entirely hypotheses on the left side and mostly evidence on the right, and the imbalance is itself information: it tells the founder where the next tests belong. Chapter 10 reads the board that way.

KEY CONCEPT The canvas is a hypothesis board, not a poster A canvas earns its page when it is versioned, dated and marked. Each item carries one of two marks — evidence, with a reference to the validation log, or hypothesis, with the test that would confirm it — and each version records what changed since the last one and why. A founder who can show three dated canvases, each one revised on evidence, has demonstrated the thing that lenders, investors and co-founders are actually looking for: not a model that is right, but a founder who finds out when it is wrong. The poster on the wall demonstrates the opposite.

6.1.3 Where the canvas sits among the frameworks

Other descriptions of a business model exist, and knowing them clarifies what the canvas does. Mark Johnson, Clayton Christensen and Henning Kagermann described a model as four interlocking elements — a customer value proposition, a profit formula, key resources and key processes — and argued that innovation in a model is a change in one element that forces changes in the others (Johnson, Christensen, & Kagermann, 2008). David Teece defined a business model as the design of the value creation, delivery and capture mechanisms, and observed that a model is easy to imitate unless it is coupled with something hard to replicate (Teece, 2010) — the unfair advantage of Chapter 8. Ash Maurya adapted the canvas for early-stage ventures as the Lean Canvas, replacing four blocks with problem, solution, key metrics and unfair advantage (Maurya, 2012); Section 6.5 says when to use it. The canvas of this chapter is the common language; the others are lenses on the same object.

6.2 The Right Side

The right side of the model — segments, propositions, channels, relationships, revenue — is the customer-facing half, and most of it has been written already. This section places the work of the previous chapters in its blocks and adds what those chapters did not cover: channels and relationships.

6.2.1 Customer segments

The block holds the segments the venture serves, and for a new venture it holds one: the beachhead of Chapter 5, with the ideal customer profile as its description and the runner-up written in the margin. Osterwalder and Pigneur distinguish five kinds of segment structure — a mass market with one broad segment, a niche market with one narrow one, a segmented market with several related groups, a diversified market with unrelated ones, and a multi-sided market with two or more interdependent groups that the model must serve together, like the restaurants and the landscapers of Norvale's case, or the buyers and sellers of a marketplace (Osterwalder & Pigneur, 2010). A founder should name which structure the model has, because the multi-sided kind changes every other block and Section 6.4 treats it as a pattern of its own.

6.2.2 Value propositions

The block holds one proposition per segment, in the statement form of Chapter 4, with a pointer to the profile and the value map behind it. When the interviews split a segment, the canvas gets two segment notes and two proposition notes, connected — and the founder discovers, looking at the page, whether the two propositions need two channels, two relationships and two revenue streams as well, which is the moment the venture decides whether it is one business or two.

6.2.3 Channels

The block answers, for each segment, how customers become aware of the offer, how they evaluate it, how they purchase it, how it is delivered to them, and how they get support after the sale — five phases that a single channel rarely covers alone (Osterwalder & Pigneur, 2010). A founder distinguishes own channels (the website, a sales conversation, a shop) from partner channels (a distributor, a marketplace, a referrer), and direct from indirect; own and direct channels keep the margin and the relationship, partner and indirect channels bring reach at a cost. For a new venture, the channels block should hold only channels that exist today — the association, the two referring accountants, the founder's network, a web page — and the beachhead criteria of Chapter 5 already required that at least one of them reach the segment. The marketing system that operates channels is Book B1's subject; the channels themselves are Collections C, D and F.

6.2.4 Customer relationships

The block answers what kind of relationship each segment expects and the model can afford: personal assistance (a human, during and after the sale), dedicated assistance (one named person per customer), self-service (the customer helps themselves), automated service (self-service with recognition and personalization), a community where customers help each other, or co-creation where customers contribute to the offer (Osterwalder & Pigneur, 2010). The relationship block is where many small-venture models break silently: a proposition that promises "a bookkeeper who knows your business" is a dedicated-assistance relationship, and it caps the number of customers one bookkeeper can serve, which is exactly the capacity limit that bounded the SOM in Chapter 5. Naming the relationship makes the cap visible.

6.2.5 Revenue streams

The block holds, for each segment, how the money comes in — a one-time sale, a usage fee, a subscription, a license, a brokerage fee, a rental, advertising — and how the price is set. Chapter 7 develops the mechanisms and the arithmetic; on the canvas, the block records the hypothesis ("$180 per session, four sessions a year, paid at each session") and the evidence behind it ("four deposits at $150; seven owners asked when"). A revenue note that cannot be connected by an arrow to a segment note is a stream from nobody, and it is the first thing a critical reader will find.

6.3 The Left Side

The left side — resources, activities, partners, costs — describes what the venture must be and do to deliver the right side. It is the half founders fill last and worst, because it is about the founder rather than the customer, and it is where feasibility failures hide.

6.3.1 Key resources

The block lists the assets without which the model does not work: physical resources (a kitchen, a vehicle, a workshop, equipment, stock), intellectual resources (a brand, a method, data, a licence, a patent, a customer list), human resources (the founder's skills, a specific hire, a network), and financial resources (working capital, a line of credit, deposits) (Osterwalder & Pigneur, 2010). The discipline is to list only what is key — what the model would stop without — and to mark each as owned, to be acquired, or to be obtained from a partner. Mariam's key resources were her own expertise and her client relationships (human and intellectual, owned), a second bookkeeper by year two (human, to be hired) and a working-capital cushion for the months between sessions (financial, to be arranged).

6.3.2 Key activities

The block lists what the venture must do well, and Osterwalder and Pigneur group activities into three families: production — making and delivering the offer; problem solving — the consulting, service and support work that solves individual customers' problems; and platform or network management — running the system that connects the sides of a multi-sided model (Osterwalder & Pigneur, 2010). The family tells the founder what kind of business they are building. A receipt clinic is problem-solving; a sorting kit is production; a matching service between owners and bookkeepers is platform management, and each family has its own cost structure and its own skills.

6.3.3 Key partners

The block lists who supplies what the venture will not do itself, and why. Partnerships exist for three reasons — to optimize and gain scale (a supplier does it cheaper), to reduce risk (a partner shares the uncertainty), and to acquire resources or activities the venture lacks (a partner has the license, the reach, the skill) — and they take several forms, from a simple buyer–supplier relationship to a strategic alliance, a partnership between competitors, or a joint venture (Osterwalder & Pigneur, 2010). For a new venture, the partners block is a list of dependencies, and each one is a risk to be read in Section 6.6: the referring accountants are a partner channel that could stop referring; the delivery company is a supplier that could raise its price; the software the whole service runs on is a partner that could change its terms. Book A8 (Vertical Integration and Make-or-Buy) treats the choice between doing and partnering in full; Collection F treats partner channels.

6.3.4 Cost structure

The block describes the costs the model incurs, in kind rather than in amounts at this stage — Chapter 7 puts the numbers in. The founder notes whether the model is cost-driven, built around the lowest cost structure and the leanest delivery, or value-driven, built around premium value with the costs that implies; which costs are fixed — rent, salaries, subscriptions, whether or not anyone buys — and which are variable — materials, transaction fees, the hours of a session, incurred per sale; and where economies of scale (unit cost falls with volume) or economies of scope (one resource serves several offers) might appear (Osterwalder & Pigneur, 2010). A canvas whose cost block says only "salaries, rent, software" has not been thought about; one that says "variable: two bookkeeper-hours and 40 km of travel per session; fixed: the software subscription and a share of the office; step cost: the second bookkeeper at 40 clients" has.

WATCH OUT Filling the left side first Founders who start from a skill or a technology fill the canvas from the left: the resources they have, the activities they enjoy, the partners they know, and only then a customer who might want what results. The canvas then describes the founder's capabilities with a customer attached, which is the solution-in-search-of-a-problem of Chapter 1 drawn on one page. The consequence is a left side full of facts and a right side full of hopes — and a venture that is feasible, perhaps viable, and not desirable. Fill the right side from evidence first; then ask, block by block, what it takes to deliver it. If the answer is a resource, an activity or a partner the founder cannot obtain, the left side has just done its job: it has found a feasibility failure on paper, at the cost of an afternoon.

6.4 Business Model Patterns

Most business models are variations on a small number of patterns — recurring arrangements of the nine blocks that have been proved to work in many industries. Osterwalder and Pigneur describe several, Oliver Gassmann and his colleagues cataloged fifty-five, and a founder who recognizes the pattern of a candidate model inherits, with it, the knowledge of what that pattern requires and where it fails (Osterwalder & Pigneur, 2010; Gassmann, Frankenberger, & Csik, 2014). Table 6.1 presents eight patterns that small ventures use most. For each pattern it states how the model makes money, what must be true for it to work, its typical failure, and a small-venture example; read the "what must be true" column as a list of hypotheses that the pattern adds to the founder's own.

Table 6.1 — Eight business model patterns for small ventures

6.4.1 Choosing a pattern

The pattern is chosen after the right side is filled, never before, and the choice is a hypothesis like any other. Three questions guide it. Does the job recur? If it does, subscription or productized service; if it happens once, a sale or a project. Does the venture serve one side or two? If two interdependent groups must both be present, the model is a platform, with the chicken-and-egg problem that Geoffrey Parker, Marshall Van Alstyne and Sangeet Choudary describe as the platform's central launch challenge — one side will not come until the other is there — and with the answers they document: subsidize one side, bring a first side yourself, start in one geography or one niche where both sides already meet (Parker, Van Alstyne, & Choudary, 2016). And can the offer be delivered by others? If the method can be written down and taught, licensing is a pattern that turns a capacity-bounded practice into a business — the option that the sizing of Chapter 5 opened for Mariam.

DEFINITION Multi-sided platform A multi-sided platform is a business model that creates value by bringing together two or more distinct but interdependent groups of customers, each of which values the platform more as the other group grows — a marketplace, a matching service, a payment network, a community with advertisers. Its defining property is the network effect: the value to each user rises with the number of users on the other side. Its defining difficulty is the start, when neither side has a reason to join; its defining risk is disintermediation, when the two sides, once matched, deal directly and bypass the platform (Parker et al., 2016). Book A9 (Platform Strategy: Integrating with Dominant Offers and Partner Ecosystems) develops the model; a first venture should adopt it only with a specific answer to how both sides arrive.

6.4.2 Free, and what pays for it

Several patterns involve giving something away, and Chris Anderson's analysis of free offers is the useful caution: "free" is never free for the venture, and a free element must be paid for by someone — the paying minority in freemium, the consumable in razor-and-blade, the advertiser in an advertising model, or the customer's own later purchase in a lead magnet (Anderson, 2009). A founder who adds a free tier must write, on the canvas, who pays for it and why they will keep paying; the arithmetic of the conversion rate belongs in Chapter 7.

CASE IN POINT Three canvases for one idea Fictional example; names and figures are invented. After the sizing of Chapter 5, Mariam drew her opportunity three ways in one afternoon. Canvas A — the service. Segment: owners above $300,000 in the region. Proposition: the quarterly clinic. Channels: the association, two referring accountants, her own clients. Relationship: dedicated — "your bookkeeper". Revenue: $720 per client per year. Resources: her expertise and relationships; a second bookkeeper. Activities: problem solving, four sessions a year per client. Partners: the accountants. Costs: bookkeeper-hours, travel, software. A productized-service pattern, capacity-bounded, obtainable market about $57,600 a year. Canvas B — the kit. Segment: bookkeeping practices across the country. Proposition: a licensed method — envelope kits, a session script, a client onboarding sequence — that lets any practice run the clinic. Channels: the bookkeepers' association newsletter, a webinar, search. Relationship: automated, with a community of licensees. Revenue: a license fee per practice per year plus kit sales. Resources: the documented method (intellectual, to be written), a website, a supplier of kits. Activities: production and community. Partners: the association, a printer. Costs: writing the method once; low variable costs. A licensing pattern, channel-bounded, obtainable market about 96 practices by year three. Canvas C — the platform. Segments: owners who need a clinic, and bookkeepers who would run one. Proposition to owners: a booked, vetted clinic near them; to bookkeepers: clients. Channels: search and the association, both sides. Relationship: automated matching. Revenue: a commission per booked session. Resources: a booking system, a supply of bookkeepers. Activities: platform management, quality control. Partners: none critical. Costs: software, marketing to both sides. A multi-sided pattern with a chicken-and-egg start that Mariam, alone and with a day job, could not fund. She chose A for year one and wrote B as the next pin — the licensing pattern that Chapter 5's next-pin criterion had suggested — and put C in the log as "not now, and not by me". Three canvases, one afternoon, three businesses; the choice was made on evidence and capacity rather than on which one sounded biggest.

6.5 Filling the Canvas

The canvas is filled in a workshop, on a wall or a whiteboard tool, with sticky notes rather than sentences, because notes can be moved, grouped and thrown away, and sentences cannot.

METHOD The canvas workshop (two hours) When to use it: once the beachhead, the ICP and the tested value proposition exist, and before the numbers of Chapter 7. Steps: (1) Draw the nine blocks large enough to hold notes; bring the validation log, the profile, the value map and the ICP. (2) Fill the right side first, in order — segments, propositions, channels, relationships, revenue — one item per note, and mark each note evidence (with the log reference) or hypothesis (with the test that would confirm it), using two colors. (3) Fill the left side by asking, for each right-side note, "what does it take to deliver this?" — resources, activities, partners — and mark each note owned, to be acquired or from a partner. (4) Fill the cost block in kinds: fixed, variable, step; cost-driven or value-driven. (5) Name the pattern of Table 6.1 the model most resembles, and add its "what must be true" as hypothesis notes. (6) Photograph or export the board, date it, and write the version note: what this version assumes that the previous one did not. Deliverable: canvas v1, dated, with every note marked, and a list of hypotheses extracted for Chapter 10. Pitfalls: writing paragraphs on notes; filling the left side first; filling one block per segment when there are two segments (use two colors of note, one per segment, on the same board); treating the canvas as finished; and the most common — marking as evidence what is actually a hypothesis the founder likes.

6.5.1 Rules of the board

Four rules keep the board honest. One item per note, so that items can be moved and removed individually. One color per segment, so that a two-segment model shows, at a glance, which blocks the segments share and which they do not. Two marks — evidence and hypothesis — on every note, so that the board is a hypothesis board. And a photograph and a date at the end of every session, so that versions can be compared; a founder who cannot say what changed between canvas v2 and v3 has not been running discovery.

6.5.2 The Lean Canvas variant

Ash Maurya's Lean Canvas keeps the customer segments, value proposition, channels, revenue and cost blocks and replaces key partners, key activities, key resources and customer relationships with four blocks better suited to a venture that does not yet know what it will build: problem (the top three problems, and the existing alternatives), solution (the top three features), key metrics (the numbers that show whether the model works) and unfair advantage (what cannot easily be copied or bought) (Maurya, 2012). The variant is designed for the earliest stage, when the founder's uncertainty is about the problem and the solution rather than about partners and resources, and for that stage it is the better tool; a founder in the first weeks of discovery may find it the natural way to record the opportunity brief of Chapter 2. The full canvas takes over when the venture must describe delivery — resources, activities, partners, relationships — which is where this chapter sits. Many founders keep both: the Lean Canvas as the problem-and-solution view, the full canvas as the operating view.

6.5.3 Two segments, one board

When the venture serves two segments — the owner and the kitchen manager, the owner and the bookkeeper in a platform — the board holds both, in two colors, and the founder reads the overlaps. Blocks where the colors mix (a shared channel, a shared resource) are economies of scope; blocks where they do not (two propositions, two relationships, two revenue streams) are the cost of serving two segments; and a board on which almost nothing is shared is two businesses drawn on one page, which is a finding, not a failure.

6.6 Reading a Canvas Critically

A filled canvas looks complete, and completeness is the danger: nine blocks with notes in each can describe a model that cannot work. Reading a canvas critically means checking that the blocks fit together, that the model does not depend on a single point that could fail, and that its hypotheses are visible.

6.6.1 Coherence tests

Table 6.2 lists seven coherence tests. Each one traces a connection between two blocks that a working model must have, states what a failure looks like on the board, and names what the founder does about it. Read the table as a checklist to run on every version of the canvas: a model passes when every test finds its connection, and the tests that fail are the next week's work.

Table 6.2 — Seven coherence tests for a canvas

6.6.2 Dependencies and single points of failure

Once the model coheres, the founder looks for what could break it, and the answer is usually a single note. One channel — an association newsletter, a referring accountant — through which most customers arrive is a dependency: if it closes, the model has no customers. One partner — a supplier, a software platform, a subcontractor — who provides a key resource is a dependency: if it fails or raises its price, the model has no delivery. One customer who provides most of the revenue is a dependency: if it leaves, the model has no money. One person — the founder — who performs every key activity is the most common dependency of all, and it is also the capacity limit of Chapter 5. The remedy is not to remove every dependency, which a new venture cannot afford, but to name each one on the board, to write what would happen if it failed, and to put the ones that would kill the model into the assumption map of Chapter 10 as risks to be reduced — a second channel, a second supplier, a documented method.

6.6.3 The "what must be true" list

The critical reading ends with a list: every hypothesis note on the board, plus every "what must be true" of the chosen pattern, plus every dependency that would kill the model, written as testable statements in the form of Chapter 1. That list is the input of Chapter 10, where the hypotheses are ranked by importance and evidence, and of Chapter 9, where the riskiest ones are tested. A canvas whose list is empty has not been read critically; a canvas whose list has thirty items has been, and thirty is normal.

WHO DOES THIS The canvas in the three situations In a solo venture the founder draws the canvas alone and reads it critically with an outsider — a mentor, an experienced owner, an accountant — because the coherence tests are easiest to pass when nobody is checking; the outsider's job is to run Table 6.2 aloud. In a founding team, the canvas workshop is done together, with one member owning the right side and another the left, and the disagreements are written on the board as hypotheses rather than settled by argument; the team also uses the board to make its own division of labor explicit — who performs which key activity. In an established business, the new offer's canvas is drawn beside the existing model's canvas, and the founder marks which notes are shared (a channel, a resource, a relationship) and which are new; the shared notes are the new line's advantage, and the new notes are what the business must learn — and the point where the existing organization's habits are most likely to smother the new model. In all three, the canvas is versioned and dated, and the "what must be true" list is what the next month tests.

QUÉBEC & CANADA NOTE Cooperatives and social enterprises as models (as of September 2026) Not every model is designed to capture value for a single owner. In Québec, a cooperative is governed by the Cooperatives Act, CQLR c C-67.2: an enterprise owned and democratically controlled by its members — consumers, producers, workers, or several of these in a solidarity cooperative — which serves its members' needs and distributes surpluses according to members' use rather than to capital. Cooperatives are a recognized model for ventures where the customers, the suppliers or the workers are the natural owners — a group of bookkeepers pooling a method, a community owning its grocery — and they have access to dedicated support and financing; the Conseil québécois de la coopération et de la mutualité and the regional cooperative development networks are the entry points. A social enterprise pursues a social mission through commercial activity; Québec's Social Economy Act, CQLR c E-1.1.1 recognizes social economy enterprises — cooperatives, mutuals and non-profit organizations with a commercial activity — and the Chantier de l'économie sociale is their network. On the canvas, these models change the segments block (members are customers and owners), the revenue block (surpluses, member contributions, grants alongside sales) and the relationships block (governance is a relationship). The legal forms themselves — cooperative, non-profit corporation, business corporation — and their registration are the subject of Book A4 (Legal Structures, Registration, Permits and Licences); the point here is that the pattern is chosen on the canvas before the form is chosen in the law.

6.7 From Canvas to Story

A canvas that only its author can read is a private document. The last step of this chapter turns it into a story of about two minutes — the length of an elevator ride, an introduction at an association meeting, or the first paragraph of a financing file — that a customer, a co-founder or a lender can follow and repeat.

6.7.1 The seven sentences

The story follows the reading order of Figure 6.1, in seven sentences. The customer: who they are, in the circumstance that makes the job urgent. The problem: what it costs them today, in their words and their numbers. The offer: what they get, and how it differs from what they do now — the value proposition statement. How it reaches them: the channel and the relationship, concretely. How the money works: what they pay, for what, how often — and, in one clause, why the numbers close. What it takes: the key resource, activity or partner that makes it possible — and, if it is the founder's own access or expertise, say so. Why us, and what is next: the unfair advantage of Chapter 8, and the next pin. Mariam's story ran: "Owners of small businesses with staff and no bookkeeping help lose a week every March and sometimes a penalty getting a year of receipts to their bookkeeper; nine of the twelve I interviewed described it, and five said they were ashamed of it. The receipt clinic sorts the year with them in four two-hour sessions at a fixed fee, so there is no March and no app to keep up. Clients come through the accountants who already refer me and the association I belong to, and each gets me as their bookkeeper. They pay $180 a session, four times a year — about $720 — for a service that costs me two hours and a drive, so it pays from the first client. It takes my fifteen years of doing this and a second bookkeeper by next year. Nobody else in the region does the sorting with the owner; four of the first seven I asked paid a deposit; and once the method is written, other practices can license it."

6.7.2 Testing the story

The story is tested like the proposition of Chapter 4: read to five people, at least two from the segment and one who lends or invests, who are asked to say back who the customer is, what they get and how the money works. Three of five should be able to; if they cannot, the sentence they stumbled on names the block that is unclear on the canvas. The story is also the founder's first pitch, and its honesty — "nine of twelve", "four deposits", "a second bookkeeper by next year" — is what distinguishes it from the pitches that lenders and investors hear every week.

Costly Mistakes

1. Treating the canvas as a poster. Filled once, framed and never versioned, it records the founder's beliefs on one day and hides every later contradiction. The consequence is a model that drifts from its evidence while its page stays reassuring — and a financing file built on the poster.
2. Filling the left side first. The canvas describes the founder's capabilities with a customer attached; the right side is hope and the left side is fact. The consequence is a feasible, perhaps viable and undesirable venture — the solution in search of a problem, drawn on one page.
3. A revenue stream from nobody. "Advertising", "commissions" or "premium tier" with no paying segment named and no evidence that it will pay. The consequence is a model whose money is a wish, found by the first critical reader — or by the bank account.
4. A relationship the revenue cannot afford. A dedicated bookkeeper promised to a $720-a-year customer, a human onboarding for a $9-a-month subscriber. The consequence is a model that loses money on every customer it delights, discovered in Chapter 7 if the canvas was honest and at the end of year one if it was not.
5. A platform without an answer to the chicken and the egg. Two sides, neither of which will come first, and a founder alone with a day job. The consequence is a marketplace with a beautiful interface and no transactions, and a year of marketing to two empty rooms.
6. Dependencies unnamed. One channel, one partner, one customer or one person on whom the whole model rests, unmarked on the board. The consequence is a model that works until the newsletter stops, the supplier raises prices or the founder gets sick — and no plan for any of the three.

Key Takeaways

• The canvas describes a business model in nine blocks on one page, filled in reading order — who and what, how it reaches them, what it takes, the money — so that the whole model can be checked against itself.
• It is a hypothesis board, not a poster: every note is marked evidence or hypothesis, every version is dated, and the founder can say what changed between versions and why.
• The right side comes from Chapters 3 to 5 — the beachhead, its profile, its proposition — plus two blocks that name how customers are reached across five phases and what kind of relationship they expect; the relationship block is where capacity limits become visible.
• The left side — resources, activities, partners, costs — is filled by asking what it takes to deliver the right side, and it is where feasibility failures are found on paper for the price of an afternoon.
• Patterns — subscription, platform, freemium, razor-and-blade, unbundling, long tail, productized service, licensing — bring their own "what must be true", and the pattern is chosen after the right side, on whether the job recurs, whether there are two sides, and whether others could deliver the method.
• A multi-sided platform is defined by its network effect and its chicken-and-egg start; a first venture adopts it only with a specific answer to how both sides arrive.
• Seven coherence tests and a search for single points of failure turn a filled canvas into a read one, and the "what must be true" list they produce is the input of Chapters 9 and 10.
• The model becomes a two-minute story in seven sentences, tested on five people who must say it back; its honesty about the evidence is what distinguishes it from a pitch.

Your Action Plan

The plan below takes about six hours over one week and produces the versioned canvas, the pattern choice and the hypothesis list that Chapters 7, 9 and 10 use. Steps 1 and 2 fill and mark the board; step 3 chooses the pattern; steps 4 and 5 read the board critically; step 6 tells the story.

In La Machine

La Machine has no canvas of its own, and it says so; what it has is the place to hold the model's blocks as living objects and the mechanics of several revenue patterns.

• Projects (/projets) hold the venture's canvas as a project whose tasks test the hypotheses, so that the board and its tests live together. A project carries a name and a description and displays no attachment, so the photographed board, the version notes, the dependency list and the hypothesis list, each dated, are attached as document links to the tasks that use them, and the project itself is loaded by whoever administers the platform.
• Notes (/presse-papier) hold the individual notes of the board when the founder prefers them as records — one note per item, colored evidence or hypothesis, attached to the project — and Links (/liens) hold the whiteboard tool where the board itself is drawn (Chapter 13).
• Packages and subscription packages (business/packages, business/subscription-packages) are the platform's mechanics for two revenue patterns: a packaged, fixed-price offer — the productized service of Table 6.1 — and a recurring subscription; a founder who chooses either pattern configures the offer here, and the revenue block of the canvas becomes a configured product rather than a note.
• The marketplace (client/marketplace) and the store (business/shop) are the platform's own multi-sided and one-sided sales surfaces: a venture that sells a packaged offer, a course or a plan to the platform's users lists it there, which is one channel note the canvas can hold with evidence.
• Arbitrages (/arbitrages) records the pattern choice and the alternative, with the reasoning, as the action plan's step 3 requires.

What La Machine does not do: it offers no graphical business model canvas or Lean Canvas, no pattern library and no coherence check — the board is drawn on a whiteboard tool from Chapter 13 or on a wall, and Table 6.2 is run by a person. Book A6 (Designing a Value-Added Offer) develops packages, tiers and guarantees; Book A9 (Platform Strategy: Integrating with Dominant Offers and Partner Ecosystems) develops the multi-sided model; Book N2 (Piloting Tasks, Executions and Projects) shows how a project groups a venture's tasks, and how each task keeps its documents and its runs.

Self-Check

1. Name the nine blocks of the canvas in the order this chapter fills them, and say why the order matters.
2. What is the difference between a canvas used as a poster and a canvas used as a hypothesis board, and what does a founder show with three dated versions?
3. Why is the relationship block where capacity limits become visible? Give an example.
4. Name four of the eight patterns of Table 6.1 and, for each, one thing that must be true for it to work.
5. Give three of the seven coherence tests and describe what a single point of failure is on a canvas.

Answers. (1) Customer segments, value propositions, channels, customer relationships, revenue streams, key resources, key activities, key partners, cost structure; each block depends on the ones before, and filling the left side first describes the founder's capabilities with a customer attached (Sections 6.1.1, 6.3). (2) A poster records beliefs on one day and is never revised; a hypothesis board marks every note evidence or hypothesis, is versioned and dated, and records what changed and why; three dated versions show a founder who finds out when the model is wrong (Section 6.1.2). (3) The relationship a proposition promises — dedicated assistance, "your bookkeeper" — determines how many customers one person can serve, which is the capacity that bounds the obtainable market for a service; naming the relationship makes the cap visible (Section 6.2.4). (4) For example: subscription — the job recurs and the cost to serve each period is well below the fee; multi-sided platform — each side values the other's presence and both can be attracted at once; freemium — the cost of a free user is near zero and a meaningful share converts; productized service — the scope can be standardized without losing the outcome (Table 6.1). (5) Any three of Table 6.2, for example: every segment has a proposition; every revenue stream comes from a named segment; every key resource is owned, to be acquired or supplied by a named partner. A single point of failure is one note — a channel, a partner, a customer or a person — on which the whole model depends and whose failure would stop it; it is named, its consequence written, and the killing ones go to the assumption map (Section 6.6.2).

Chapter References

Cooperatives Act, CQLR c C-67.2. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/C-67.2

Social Economy Act, CQLR c E-1.1.1. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/E-1.1.1

Anderson, C. (2009). Free: The future of a radical price. Hyperion.

Gassmann, O., Frankenberger, K., & Csik, M. (2014). The business model navigator: 55 models that will revolutionise your business. Pearson.

Johnson, M. W., Christensen, C. M., & Kagermann, H. (2008). Reinventing your business model. Harvard Business Review, 86(12), 50–59.

Maurya, A. (2012). Running lean: Iterate from plan A to a plan that works (2nd ed.). O'Reilly Media.

Osterwalder, A., & Pigneur, Y. (2010). Business model generation: A handbook for visionaries, game changers, and challengers. Wiley.

Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform revolution: How networked markets are transforming the economy and how to make them work for you. W. W. Norton.

Strategyzer AG. (n.d.). The Business Model Canvas. Retrieved September 7, 2026, from https://www.strategyzer.com/library/the-business-model-canvas

Teece, D. J. (2010). Business models, business strategy and innovation. Long Range Planning, 43(2–3), 172–194.

Chapter 7 — Revenue Streams, Costs and Unit Economics

The canvas says "$180 a session, four times a year", and it says "bookkeeper-hours, travel, software". Between those two notes lies the question that decides whether the venture is a business: does one customer, served once, leave more money than it cost to win and serve — and how many customers does it take before the fixed costs are covered and the founder is paid? This chapter answers it with first estimates, in arithmetic a founder can do in an evening, and then asks what happens when the estimates are wrong.

What You'll Be Able to Do

• Choose how the model makes money among seven revenue mechanisms, and know when the cash arrives for each.
• Form a first pricing hypothesis from the value the customer described and the alternatives they use, and test it without a spreadsheet of wishes.
• Separate fixed, variable and step costs, include the founder's own time, and keep sales taxes out of revenue.
• Compute the unit economics — contribution margin, customer acquisition cost, lifetime value, payback — for one customer, and read the ratios between them.
• Compute the break-even for a service and for a product, and sketch the first twelve months.
• Tell cash from profit, compute burn and runway, and know how long the venture can survive.
• Stress-test the numbers with sensitivities and scenarios, and write the three numbers that must be true.

Why It Matters

Unsustainable unit economics was cited in about a fifth of the start-up post-mortems of Chapter 1, and running out of capital in seven out of ten; the two are the same failure seen from the profit line and from the bank account. A venture that loses money on every customer cannot be saved by more customers, and a venture that makes money on every customer can still die waiting for the cash to arrive. For the founder, the arithmetic of this chapter is the difference between a business that pays them and a hobby that costs them. For a team or an established business, it is the page that turns "we believe in this" into "this works at forty customers, and here is what must be true."

In This Chapter

• 7.1 How Models Make Money
• 7.2 First Pricing Hypotheses
• 7.3 Cost Structure
• 7.4 Unit Economics
• 7.5 Break-Even and First Estimates
• 7.6 Cash: Timing, Working Capital and Runway
• 7.7 Stress-Testing the Numbers

7.1 How Models Make Money

The revenue block of the canvas holds a mechanism — the way in which value delivered becomes money received — and a founder should choose it consciously, because each mechanism has its own timing, its own predictability and its own failure. Osterwalder and Pigneur list the main ones: the sale of an asset, a usage fee, a subscription fee, lending, renting or leasing, licensing, a brokerage fee and advertising (Osterwalder & Pigneur, 2010). Table 7.1 presents seven of them. For each mechanism it states what the customer pays for, when the cash typically arrives, how predictable the revenue is, a small-venture example and the pitfall that most often breaks it. Read the "when the cash arrives" column with particular care: it is the difference between a model that finances itself and a model that needs financing, which Section 7.6 develops.

Table 7.1 — Seven revenue mechanisms

7.1.1 One-time and recurring

The most important distinction among the mechanisms is whether revenue restarts at zero every period or carries over. A product sale, a project and a commission are one-time: each customer must be won again, or a new one found, and the venture's revenue next month is the revenue it earns next month. A subscription, a license and a maintenance contract are recurring: the customers won last year pay again this year unless they leave, and the venture starts each period with a base. Recurring revenue is more valuable per dollar — lenders, buyers and investors price it higher — and it is harder to win, because the customer commits to a relationship rather than to a purchase. Many small ventures combine the two: a one-time setup and a recurring service, or a product and its consumables (the razor-and-blade pattern of Chapter 6).

7.1.2 Fixed and dynamic pricing mechanisms

Within a mechanism, the price itself can be fixed — a list price, a price by feature or tier, a price by segment, a price by volume — or dynamic — negotiated per deal, adjusted with demand or time, set by an auction. A new venture almost always uses fixed prices, because a fixed price can be written on a page and tested, and because dynamic pricing requires data the venture does not have. The tiers, the discounts and the negotiation rules are the subject of Book A7 (Pricing Strategy and Revenue Models); this chapter needs one price hypothesis per segment, which Section 7.2 forms.

7.1.3 When the cash arrives

A mechanism also fixes when the money comes in, and that timing is a design choice as much as the price. A subscription paid in advance finances the venture with its customers' money; a project billed on delivery with 45-day terms finances the customer with the venture's money. A founder chooses the mechanism partly for its cash timing — deposits, advance payment, short terms — within the rules that Chapter 3 gave for consumer prepayments, and Section 7.6 shows what the choice is worth.

7.2 First Pricing Hypotheses

Pricing is a discipline of its own, and Book A7 gives it a full treatment. At the discovery stage the founder needs one defensible hypothesis per segment, formed from evidence rather than from cost or from fear, and a way to test it that does not rely on asking people what they would pay.

7.2.1 Three ways to price, and the one to start from

Prices are set in three ways. Cost-plus starts from what the offer costs and adds a margin; it guarantees a margin per unit if the volume arrives, and it ignores the customer entirely. Competition-based starts from what the alternatives charge and positions the offer above, at or below; it is realistic about the market and it assumes the alternatives priced correctly. Value-based starts from what the offer is worth to the customer — the cost of the pain removed, the value of the gain created — and sets a price that leaves the customer a clear surplus; it is the method the value proposition of Chapter 4 makes possible, and it is where a founder should start, checking the result against the other two.

7.2.2 The value-based estimate

The customer profile already contains the numbers. Mariam's interviews put the cost of the March problem at about four hours of the owner's time ($180 at $45 an hour), a late-filing penalty in some years ($450, in two of twelve cases), the bookkeeper's three unbilled hours per client ($180 at her rate) and a week of stress that nobody prices. The value removed is on the order of $400 to $800 a year per client, before the emotional job. A value-based price leaves the customer most of that surplus: $720 a year takes a large share of the measurable value and relies on the unmeasured part — the shame, the dread — to justify it, which the interviews suggest it does. That is a hypothesis with a mechanism behind it, and it can be tested.

7.2.3 Anchors and alternatives

The competition-based check asks what the customer's alternatives cost. A part-time assistant at $25 an hour for twelve hours a quarter costs $300 a quarter, or $1,200 a year, plus the finding and managing; an app costs $15 a month and the habit nobody keeps; the shoebox costs nothing and a week. The clinic at $180 a quarter sits below the assistant and above the app, which is the position the value map described. When the alternative is "nothing", the anchor is the cost of the problem itself, and the founder's job in the sales conversation of Book E1 is to make that cost visible before the price is mentioned.

7.2.4 Testing the price

Stated willingness to pay is the bottom rung of the evidence ladder, and the founder should not survey it directly. Three instruments work better. The deposit or the pre-order at the hypothesized price is the top rung: four deposits at $150 are evidence about $150. Two prices on two pages — a landing page test of Chapter 9 that shows different prices to different visitors and measures who clicks and who signs up — is behavioral evidence about relative demand. And, when a survey is used, the price sensitivity meter of Peter Van Westendorp asks four questions — at what price would the offer be so cheap that you would doubt its quality, at what price a bargain, at what price expensive but worth considering, and at what price too expensive to consider — and reads the acceptable range from where the answers cross (Van Westendorp, 1976). It measures perception rather than behavior, and it is useful for setting the range that the behavioral tests then probe.

KEY CONCEPT Revenue is a hypothesis; cash is a fact Everything in the revenue block — the price, the number of customers, the frequency, the renewal — is a hypothesis until customers have paid, and the numbers that follow in this chapter inherit that status: the unit economics, the break-even and the forecast are arithmetic on hypotheses. Cash is different. Cash is what is in the account, and it is the only number in this chapter that cannot be argued with. A founder who keeps the distinction — revenue and margin as hypotheses to test, cash as the fact that sets the clock — makes better decisions than one who treats the forecast as a fact and the bank balance as a detail.

7.3 Cost Structure

The cost block of the canvas said what kinds of costs the model incurs; this section puts structure on them, because the structure — not the total — decides how the business behaves as it grows.

7.3.1 Fixed, variable and step costs

Fixed costs are incurred whether or not anyone buys: rent, insurance, software subscriptions, a salary, a loan payment. Variable costs are incurred with each unit sold: the materials in a product, the hours of a session, the fuel of a pickup, the payment processor's fee, the shipping. Step costs are fixed within a range and jump when the range is exceeded: the second bookkeeper at forty clients, the second truck at two hundred stops, the bigger kitchen. Cost accounting treats the distinction as the basis of cost–volume–profit analysis, the method behind every break-even calculation (Datar & Rajan, 2021). A founder classifies every cost on the canvas into one of the three, because the classification determines the contribution margin, the break-even and the shape of profit as volume grows.

The classification also reveals the model's character. A model with high fixed costs and low variable costs — software, a facility, a route with a truck — loses money at low volume and makes it fast at high volume; it needs scale and financing. A model with low fixed costs and high variable costs — a service delivered by the founder's hours, a product bought and resold — makes a little money from the first customer and never makes a lot per unit; it needs margin and price. Most first ventures should be the second kind, and become the first kind only once the demand is proved.

7.3.2 The founder's time is a cost

The most common error in a first cost structure is a fixed cost of zero for the person doing most of the work. A founder who delivers the sessions, does the marketing and keeps the books has a cost — the salary they are not paying themselves, or the income from the job they are not doing — and a model that only works because that cost is zero is not a model; it is a subsidy the founder pays until they cannot. The discipline is to include an imputed salary for the founder in the fixed costs — at least what the founder would have to pay someone to do the work, and ideally what the founder needs to live — and to compute the break-even with it. The Do the Math box in Section 7.5 shows how much the number changes.

7.3.3 Direct and indirect, cost of sales and operating expenses

Two further distinctions matter for reading the numbers later. Direct costs can be traced to a unit or a customer — the session's hours, the kit's printing — and indirect costs cannot — the office, the insurance, the accountant; the direct costs are mostly variable, and the indirect mostly fixed, but not always. And accountants group costs into the cost of sales (or cost of goods sold), the direct costs of producing what was sold, and operating expenses, everything else; the difference between revenue and cost of sales is the gross margin, which lenders and buyers read first. Book G1 (Accounting Fundamentals for Business Owners) reads the statements; this chapter needs only the classification.

DEFINITION Contribution margin The contribution margin is the revenue of a unit — a session, a kit, a client, a month — minus the variable costs of that unit; it is what the unit contributes toward covering the fixed costs and, once they are covered, toward profit. Expressed as a share of the price, it is the contribution margin ratio. A unit with a negative contribution loses money every time it is sold, and no volume can repair it; a unit with a positive contribution repays the fixed costs one unit at a time, which is what the break-even of Section 7.5 counts.

WATCH OUT Counting sales taxes as revenue A price of $180 that the customer pays as $206.96 in Québec — $180 plus 5% GST and 9.975% QST — produces $180 of revenue, not $206.96. The taxes collected belong to the governments and are remitted; a venture that records them as revenue overstates its sales by about 15%, spends money that is not its own, and meets the remittance deadline with an empty account. The same error appears in the other direction when taxes paid on purchases are counted as costs: a registered business recovers most of them as input tax credits and refunds. The rule is simple and the Québec and Canada Note below gives the thresholds: prices, revenue and costs are computed before tax; taxes are tracked separately, collected in a separate mental account, and remitted on time.

QUÉBEC & CANADA NOTE GST and QST for a new venture (as of September 2026) In Québec, most sales of goods and services are subject to the federal goods and services tax (GST) at 5% under the Excise Tax Act, R.S.C. 1985, c. E-15 and to the Québec sales tax (QST) at 9.975% under the Act respecting the Québec sales tax, CQLR c T-0.1, both administered by Revenu Québec. A business must register for both once its total worldwide taxable supplies — sales, rentals, exchanges and the like, including those of its associates — exceed $30,000 in a given calendar quarter or in the four preceding calendar quarters; below that threshold it is a small supplier and may choose to register voluntarily, which allows it to recover the taxes paid on its purchases as input tax credits and input tax refunds (Revenu Québec, 2026). A registered business charges the taxes on its taxable sales, shows its registration numbers on its invoices, files returns at the frequency assigned to it and remits the net amount; the taxes collected are never its revenue. In other provinces, the GST applies alone or as part of a harmonized sales tax at rates that vary by province, administered by the Canada Revenue Agency (Canada Revenue Agency, 2026). For a founder, three practical points follow: price and forecast before tax; decide early whether to register voluntarily, because a venture with significant purchases benefits from the credits; and open a separate account, or at least a separate line in the cash sheet, for the taxes collected. Book G5 (Business Taxes in Québec and Canada) gives the full regime; the registration steps are in Book A4.

7.4 Unit Economics

The unit economics answer the question at the head of this chapter: does one customer leave more money than it costs to win and serve? They are computed for one unit — one customer over a year is the most useful unit for most small ventures; one transaction is the unit for a marketplace or a product — and they produce four numbers that every later decision reads.

DEFINITION Unit economics The unit economics of a business model are the revenue, the variable costs, the contribution and the acquisition cost of a single unit — typically one customer — and the relationships between them: the contribution margin, the customer acquisition cost, the lifetime value and the payback period. They describe whether the model makes money at the level of one customer, independently of the fixed costs and the volume, and they are the first numbers a founder computes and the first numbers a lender or an investor asks for.

7.4.1 Contribution per unit

The contribution per unit is the price minus the variable costs, as Section 7.3 defined it. For a customer-year it is the contribution per transaction times the number of transactions per year. Figure 7.1 draws it as a waterfall for the receipt clinic: a session priced at $180 carries $95 of variable cost — two bookkeeper-hours at a loaded cost of $40 and $15 of kit and travel — and contributes $85, a 47% margin; four sessions contribute $340 per client per year.

Figure 7.1 — The unit-economics waterfall. A session priced at $180 minus $95 of variable costs contributes $85; four sessions a year contribute $340 per client; the $150 cost of acquiring the client leaves $190 in the first year, and the acquisition is repaid after about two sessions. Original figure; illustrative figures from the fictional receipt-clinic case.

7.4.2 Customer acquisition cost

The customer acquisition cost (CAC) is everything spent to win customers over a period, divided by the number of customers won in that period — and "everything" includes the founder's hours at their imputed rate, not only the money paid to a platform or an event. A CAC computed from cash alone, when the founder spent twenty hours a month selling, understates the true cost by the value of those hours and produces a model that stops working the day the founder's time is no longer free. The CAC is computed by channel wherever possible, because channels differ by a factor of ten and the channel decision of Chapter 5 depends on it.

7.4.3 Lifetime value

The lifetime value (LTV) is the contribution a customer produces over the whole relationship: the contribution per year times the expected number of years. For a recurring model, the expected lifetime is estimated from the renewal rate — if 80% of customers renew each year, the average customer stays about 1 ÷ (1 − 0.80) = 5 years — and the estimate is a hypothesis until a year has passed. Two cautions keep the LTV honest. It is computed on contribution, not on revenue: a customer who pays $720 a year and costs $380 to serve is worth $340 a year, not $720. And it is discounted in any serious model, because a dollar five years away is worth less than a dollar today; at the discovery stage an undiscounted LTV with a conservative lifetime — three years rather than five — is an acceptable first estimate, stated as such.

7.4.4 Payback and the ratios

The payback period is the time it takes a customer's contribution to repay the cost of acquiring them: CAC divided by contribution per month. It is the number that connects unit economics to cash, because a venture must finance every customer until payback, and a long payback with many customers is a large financing need — a paradox that Section 7.6 explains. The LTV-to-CAC ratio summarizes the whole: a customer worth three times what they cost to acquire leaves room for the fixed costs, the founder's salary and the errors in the estimates; a ratio near one means the venture spends its future to buy its present. Practitioners of subscription software use a ratio above 3 and a payback under twelve months as rules of thumb (Skok, 2013); for a service business with a short payback the ratio matters less than the absolute contribution, and for a one-time product the "lifetime" is one purchase and the ratio collapses to the contribution per sale over the cost per sale. The rules of thumb are benchmarks from one industry, not laws.

DO THE MATH CAC, LTV and payback for the clinic Illustrative figures, consistent with Figure 7.1. Acquisition. In her first quarter, Mariam sponsors an association breakfast ($600) and spends twenty hours on outreach, calls and first conversations (at her imputed rate of $60, $1,200), for a total acquisition spend of $1,800; twelve clients sign. CAC = $1,800 ÷ 12 = $150 per client. By channel: the eight clients who came through the association cost $1,200 ÷ 8 = $150 each, and the four who came from the two referring accountants cost only her hours, about $60 each — which tells her where the next dollar goes. Contribution. $85 per session, $340 per client-year (Figure 7.1). Payback. $150 ÷ $85 = 1.8 sessions; at one session a quarter, about five months. Lifetime value. If 80% of clients renew each year, the expected lifetime is 1 ÷ 0.20 = 5 years and the undiscounted LTV is $340 × 5 = $1,700; at a conservative 60% renewal, the lifetime is 2.5 years and the LTV is $850. The ratio. LTV ÷ CAC = $1,700 ÷ $150 ≈ 11 at 80% renewal, and $850 ÷ $150 ≈ 5.7 at 60% — comfortable either way, because the payback is short and the channel is cheap. The renewal rate is the hypothesis that matters, and it cannot be known before the first March has passed; the log records it as such.

7.5 Break-Even and First Estimates

The unit economics say whether one customer pays. The break-even says how many customers it takes before the fixed costs are covered — the number at which the venture stops losing money — and the first estimates put that number on a calendar.

7.5.1 The break-even

The break-even in units is the fixed costs of a period divided by the contribution per unit in that period; in revenue, it is the fixed costs divided by the contribution margin ratio. The arithmetic is trivial; the judgment is in what counts as fixed — the founder's imputed salary above all — and in whether the contribution per unit holds at the volume the break-even requires (Datar & Rajan, 2021). The Do the Math box computes it twice for the clinic and once for the kit, and the difference between the first two is the most important lesson of this chapter.

DO THE MATH Break-even for a service and for a product Illustrative figures. The clinic, without the founder's time. Fixed costs of the clinic line: software $600 a year, marketing $1,200, a share of insurance and office $400 — $2,200 a year. Contribution per client-year: $340. Break-even = $2,200 ÷ $340 = 6.5, so seven clients cover the line's fixed costs. That number is the one founders compute, and it is the one that hides the truth. The clinic, with the founder's time. Mariam runs the line's administration, marketing and scheduling in about two hours a week, fifty weeks a year, at her imputed rate of $60: $6,000 — and if she delivers the sessions herself rather than through an employee, the $80 of bookkeeper-hours in the variable cost is also her time, which she must pay herself. Fixed costs with the imputed salary: $2,200 + $6,000 = $8,200. Break-even = $8,200 ÷ $340 = 24.1, so twenty-five clients — more than three times the first answer, and close to the thirty of the first-year target of Chapter 5. The line is a business at forty clients and a hobby at ten. The kit. A sorting kit sold to bookkeepers at $49, with $18 of printing, shipping and payment fees, contributes $31. Fixed costs: the design, amortized over two years, $2,000 a year; the website and the store $600; marketing $3,000; the founder's time to run it, four hours a month at $60, $2,880 — $8,480 a year. Break-even = $8,480 ÷ $31 = 274 kits a year, or 23 a month. The two models break even at similar fixed costs and very different volumes — 25 relationships or 274 transactions — which is the difference between a practice and a product, made visible in one line of arithmetic.

7.5.2 The first twelve months

A break-even is a threshold; the first estimates put it on a calendar. The founder sketches the first twelve months in a table — customers won each month from the weekly number of Chapter 5 and the ramp of its first-year target, the revenue those customers produce, the variable costs, the fixed costs including the imputed salary, and the resulting monthly result — and reads three things from it: the month in which the venture breaks even, the cumulative loss until then, and the volume at month twelve. The table is not a forecast in the sense of Book A2; it is a first estimate, and its purpose is to show the shape — how deep the trough is and how long it lasts — before the venture commits to it.

7.5.3 The assumptions table

Every number in the sketch rests on an assumption, and Table 7.2 is where the assumptions live. Each row is one assumption — a price, a frequency, a cost, a rate — with its value, the range within which it is plausible, its source or its mark (sourced, estimated or guessed, as in Chapter 5), and the test that would confirm it. The example rows are the clinic's. Read the table as the bridge between this chapter and Chapter 10: the assumptions with wide ranges and weak sources are the ones the assumption map will rank first.

Table 7.2 — The assumptions table (example rows from the receipt-clinic case)

7.6 Cash: Timing, Working Capital and Runway

A venture can be profitable on paper and dead in the bank. The reason is timing: revenue is recognized when the work is done, cash arrives when the customer pays, costs are incurred when the work is done, and cash leaves when the bills are due — four different moments. This section gives the vocabulary and the arithmetic of the gap.

7.6.1 Cash is not profit

Three mechanisms separate the two. Receivables: a business customer who pays in 45 days has received the service, and the venture has recognized the revenue, but the cash is 45 days away; a venture with $10,000 of monthly sales on 45-day terms has $15,000 permanently owed to it. Inventory and advance costs: a product bought or made before it is sold ties cash up until the sale; a course built before it is sold ties up months of the founder's time. Deposits and prepayments: a subscription paid in advance is cash the venture holds before it has earned the revenue — favorable to cash, and a liability until the service is delivered. Add the sales taxes collected and not yet remitted, and the account balance can be higher or lower than the profit by amounts that decide whether payroll is met.

7.6.2 Working capital

Working capital is the money a venture must keep tied up to operate: the receivables it waits for, the inventory it holds, the deposits it has paid, minus the payables it has not yet paid and the deposits it has received. A model's working-capital need is a design choice: a clinic paid at each session needs almost none; a kit business that prints a thousand kits before selling one needs the cost of the print run; a consulting practice that bills on delivery with 45-day terms needs a month and a half of costs in the bank. Business Development Bank of Canada's guides on cash flow are a practical starting point for a founder who has never built a cash sheet (Business Development Bank of Canada, 2026b); Book G2 (Cash Flow Management and Forecasting) treats the subject in full.

DEFINITION Runway The runway is the number of months a venture can continue at its current rate of net cash consumption before its cash reaches zero: cash on hand divided by the monthly net burn, where the net burn is the cash going out each month minus the cash coming in. It is the founder's clock. A runway is computed for the venture and, honestly, for the founder — whose personal cash and personal monthly need are part of the same equation when the venture does not yet pay a salary — and it is recomputed every month, because it shortens faster than it lengthens.

DO THE MATH Runway and burn Illustrative figures. The venture alone. Monthly cash out: $1,200 of tools, marketing and insurance in the first months; monthly cash in: nothing in month one, then the first sessions. Net burn in month one: $1,200; with $6,000 set aside, the venture's runway is 5 months before revenue, and it lengthens as sessions are paid. The founder with the venture. The founder has $18,000 of savings, needs $3,200 a month to live, and has left the day job. Total monthly burn: $3,200 + $1,200 = $4,400; runway = $18,000 ÷ $4,400 = 4.1 months — before the twenty-fifth client, before the first March, before the renewal rate is known. The founder who keeps the job. The salary covers the $3,200; the burn is the venture's $1,200; runway = $18,000 ÷ $1,200 = 15 months, which is long enough to reach break-even at the ramp of Chapter 5 and to learn the renewal rate. Same venture, same savings, one decision — and the discovery quarter of Chapter 1 assumed the second. The arithmetic is also why a long payback with many customers is a financing need: at a $150 CAC and a five-month payback, every client won ties up $150 for five months, and thirty clients won in a quarter tie up $4,500 — which the founder's runway must absorb, or a lender must (→ see Book A5, Financing a Business).

7.6.3 When the venture pays the founder

The founder's salary is a fixed cost in the break-even and a personal need in the runway, and the two meet at the month when the venture can pay it. That month is later than founders expect — after break-even on the imputed salary, not on the cash costs alone — and knowing it is the difference between a planned transition from the job and a forced one. The sketch of Section 7.5.2 gives the month; the runway says whether the founder can wait for it.

7.7 Stress-Testing the Numbers

Every number in this chapter is an estimate, and most estimates are optimistic. Stress-testing asks what happens when they are wrong, and produces the three or four numbers that must be true for the model to work — the thresholds that Chapter 10's tests will probe.

7.7.1 Sensitivities

A sensitivity changes one assumption at a time by a plausible amount — typically 20% — and reads the effect on the break-even, the payback and the twelve-month result. Four assumptions usually dominate: the price, the volume (contacts and conversion), the variable cost and, for recurring models, the renewal rate. In the clinic, a 20% lower price ($144) cuts the contribution per session to $49 and raises the break-even with the imputed salary from 25 to 42 clients — the model still works, at a volume the region can supply; a 20% higher variable cost ($114) has a similar effect; a renewal rate of 55% instead of 80% halves the lifetime value and makes acquisition the permanent activity of the practice. The sensitivities are ranked by the size of their effect, and the top one is the model's most important number.

7.7.2 Three scenarios

A scenario changes several assumptions together into a coherent story: a pessimistic case (lower price, slower ramp, higher cost, lower renewal), a base case (the estimates as written) and an optimistic case. The founder reads the pessimistic case for one thing — does the venture survive it? — because the base case is what founders plan for and the pessimistic case is what happens to about half of them. A venture whose pessimistic case exhausts the runway before break-even needs more runway, a cheaper model or a different first segment before it starts, not after.

7.7.3 The three numbers that must be true

The stress test ends in a short list: the assumptions whose plausible range crosses the line between a model that works and one that does not. For the clinic they are the price (above about $150, the model works at a volume the region can supply), the renewal rate (above about 60%, acquisition is a start-up cost rather than a permanent one) and the contacts per week (about eight, the weekly number of Chapter 5, sustained). Each is written as a threshold, each is a hypothesis in the assumptions table, and each gets a test in Chapters 9 and 10. A model with more than five such numbers is fragile; a founder who cannot name any has not stress-tested it.

CASE IN POINT Unit economics that depend on the route Fictional example; names and figures are invented. Norvale Compost (Chapter 4) priced its restaurant service at $180 a month for twice-weekly pickup — about 8.6 pickups a month, or $21 per stop. The founders' first unit economics looked at the truck: a route costs about $320 a day (a driver's four hours at $30, and $200 of fuel, insurance, maintenance and depreciation on the truck). With the ten customers of the first month scattered across the city, a route made ten stops, and each cost $32 — the venture lost $11 on every pickup, before the processing site and the founders' time. The arithmetic did not condemn the model; it condemned the geography. At twenty-five stops on the same route, the cost per stop falls to $12.80 and each pickup contributes $8.20 — about $70 a month per customer, from which the site and the founders are paid. Twenty-five stops in a four-hour route is only possible if the customers are within a few blocks of each other. The founders therefore redrew their beachhead (Chapter 5) from "restaurants in the city" to "restaurants in two adjacent commercial streets", where forty candidates stood within a kilometre, and they refused customers elsewhere for a year, however eager. Their three numbers that must be true became: stops per route above twenty, churn below 15% a year (a lost customer breaks the density) and a compost price to landscapers above $40 a tonne (the second revenue stream, which turned a break-even service into a profitable one). Two of the three were geography and retention, not price — which the unit economics found and a spreadsheet of city-wide averages would have hidden.

WHO DOES THIS The numbers in the three situations In a solo venture the founder computes everything in this chapter in a spreadsheet, and the discipline is to include the imputed salary and to have the assumptions table read by an accountant or an experienced owner, whose first question — "what did you assume for your own time?" — is usually the right one. In a founding team, one member owns the numbers, the others own the assumptions in their area (the customer member the price and the conversion rates, the product member the variable costs), and the team runs the stress test together, because the pessimistic scenario is an argument nobody wants to have alone. In an established business, the new line's economics are computed on their own — with a fair share of the shared costs and a real price for the resources borrowed from the existing business — because a new line that looks profitable only when the old business pays its rent is a hobby the company has adopted; the intrapreneur also computes the working capital the line will tie up, which the company's finance function will ask about first. In all three, the three numbers that must be true are written down and reviewed monthly.

Costly Mistakes

1. A break-even computed without the founder's time. Seven clients instead of twenty-five; the venture "works" from month three and never pays its founder. The consequence is a year of subsidy discovered when the savings run out, and a model that collapses the day the founder's hours are no longer free.
2. A price set from cost or from fear. Cost-plus ignores the customer, and fear sets the price below the value the interviews measured; the contribution per unit is too small to cover acquisition and fixed costs at any reachable volume. The consequence is a business that is busy and poor, and a price increase that is harder to make on a hundred customers than on the first ten.
3. Sales taxes counted as revenue, or credits missed. Sales overstated by about 15%, money spent that belongs to the governments, and a remittance deadline met with an empty account — or taxes paid on purchases never recovered. The consequence is a cash crisis with a tax bill attached, and interest and penalties on top.
4. Lifetime value computed on revenue and on a guessed lifetime. A $720-a-year customer valued at $3,600 over five years, when the contribution is $340 and the renewal is unknown. The consequence is a customer acquisition budget three times what the model can afford, spent before the first renewal season says the truth.
5. Profit mistaken for cash. A profitable forecast, 45-day terms, a print run bought in advance, and a payroll missed in month four. The consequence is a venture that dies solvent — or a founder who quit the job with four months of runway before a twelve-month ramp.
6. A model with no numbers that must be true. Estimates presented as facts, no sensitivities, no pessimistic case. The consequence is a model that cannot be tested, and a founder who learns which assumption was wrong only when the cash is gone.

Key Takeaways

• The revenue mechanism is a design choice with three properties: what the customer pays for, when the cash arrives, and how predictable the revenue is; recurring revenue is worth more and is harder to win, and the cash timing decides whether the model finances itself.
• The first price is value-based, formed from the cost of the pain and the value of the gain in the customer profile, checked against the alternatives and against cost, and tested by deposits, two-price pages and the price sensitivity meter rather than by asking.
• Costs are fixed, variable or step, and the founder's time is a fixed cost; a model that works only because that cost is zero is a subsidy, not a model.
• Sales taxes are never revenue. In Québec, GST at 5% and QST at 9.975% are collected for the governments once the $30,000 small-supplier threshold is crossed, and the credits on purchases are recovered by registered businesses; prices, revenue and costs are computed before tax.
• Unit economics — contribution, CAC, LTV, payback — say whether one customer pays; the LTV is computed on contribution with a conservative lifetime, the CAC includes the founder's hours and is computed by channel, and the ratios are read as benchmarks, not laws.
• The break-even is the fixed costs divided by the contribution per unit, and it triples when the imputed salary is included; a service breaks even in relationships, a product in transactions, and the difference is the difference between a practice and a company.
• Cash is not profit: receivables, inventory, prepayments and taxes separate the two; working capital is a design choice, and the runway — cash divided by monthly net burn, computed for the venture and for the founder — is the clock that every decision runs against.
• Stress-test with sensitivities and three scenarios, and write the three numbers that must be true — the thresholds at which the model stops working — because they are the hypotheses that Chapters 9 and 10 must test first.

Your Action Plan

The plan below takes about eight hours over one week and produces the numbers that Chapters 9, 10 and 14 will test and that Book A2's plan will inherit. Steps 1 and 2 form the revenue and cost hypotheses; steps 3 and 4 compute the unit economics and the break-even; step 5 builds the cash view; step 6 stress-tests and writes the thresholds.

In La Machine

La Machine provides the calculators for the arithmetic of this chapter and the mechanics that turn a revenue hypothesis into configured prices; the sheets themselves live in a spreadsheet attached to the project.

• Calculator & Tools (productivity/tools) offers, under its finance sections, the calculators this chapter uses: return on investment, gross and net margin, break-even and the point at which costs are covered, the payback period, and — for the financing questions of Section 7.6 — net present value and internal rate of return, loan amortization and the time value of money; a tax calculator handles the GST and QST arithmetic of the Québec and Canada Note. Each calculator states the formula it applies and returns a result the founder can paste into the sheet.
• Promotions (business/promotions) hold the introductory offers that the switching test of Chapter 4 may require — a first month free, a credit — as configured rules rather than improvised discounts, so that their cost appears in the acquisition cost.
• Plans (plans) and the public pricing page (/pricing) are the platform's own tiered subscription, which a founder can read as one worked example of a recurring mechanism with fixed tiers — described here, not recommended.

What La Machine does not do: it holds no financial model, no assumptions table and no cash-flow sheet — the twelve-month sketch, the scenarios and the runway are computed in a spreadsheet and attached to the project — and it does not compute lifetime value or acquisition cost from the venture's data at this stage. Book A2 (Chapter 7) builds the full financial projections, Book G1 reads the statements, Book G2 manages cash, Book G5 handles the business taxes, and Book A7 develops pricing beyond the first hypothesis.

Self-Check

1. Name four revenue mechanisms and, for each, when the cash typically arrives.
2. Why must the founder's time be included in the fixed costs, and what happens to the break-even of the clinic when it is?
3. Define contribution margin, customer acquisition cost, lifetime value and payback period, and say on what number the lifetime value must be computed.
4. Give the GST and QST rates in Québec and the threshold above which a business must register (as of September 2026), and explain why taxes collected are not revenue.
5. What is the runway, and why does it differ so much for a founder who keeps a day job? What are "the three numbers that must be true"?

Answers. (1) For example: a product sale — at delivery or on 30- to 60-day terms; a usage fee — after use, billed periodically; a subscription — in advance at the start of each period; licensing — an annual fee in advance or a royalty after the licensee's sales; a commission — at the transaction (Table 7.1). (2) Because a model that works only when the person doing most of the work is unpaid is a subsidy, not a model; with a $6,000 imputed salary, the clinic's break-even rises from seven clients to twenty-five (Sections 7.3.2, 7.5.1). (3) Contribution margin — revenue of a unit minus its variable costs; CAC — all acquisition spending, including the founder's hours, divided by customers won; LTV — the contribution a customer produces over the relationship, contribution per year times expected lifetime; payback — CAC divided by contribution per month; the LTV is computed on contribution, never on revenue (Section 7.4). (4) GST 5% and QST 9.975%; registration is mandatory once taxable supplies exceed $30,000 in a calendar quarter or the four preceding quarters; the taxes collected belong to the governments and are remitted, so recording them as revenue overstates sales and spends money that is not the venture's (Section 7.3, Québec and Canada Note). (5) The runway is cash on hand divided by monthly net burn, the months until cash reaches zero; a founder who keeps the job removes their personal need from the burn, lengthening the runway from about four months to about fifteen in the example; the three numbers that must be true are the assumptions whose plausible range crosses the line between a model that works and one that does not — for the clinic, the price, the renewal rate and the weekly contacts — written as thresholds and tested first (Sections 7.6, 7.7).

Chapter References

Act respecting the Québec sales tax, CQLR c T-0.1. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/T-0.1

Excise Tax Act, R.S.C. 1985, c. E-15. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/E-15/

Business Development Bank of Canada. (2026b). Articles and tools: Money and finance. Retrieved September 7, 2026, from https://www.bdc.ca/en/articles-tools/money-finance

Canada Revenue Agency. (2026). GST/HST for businesses. Government of Canada. Retrieved September 7, 2026, from https://www.canada.ca/en/revenue-agency/services/tax/businesses/topics/gst-hst-businesses.html

Datar, S. M., & Rajan, M. V. (2021). Horngren's cost accounting: A managerial emphasis (17th ed.). Pearson.

Osterwalder, A., & Pigneur, Y. (2010). Business model generation: A handbook for visionaries, game changers, and challengers. Wiley.

Revenu Québec. (2026). Registering for the GST and QST. Gouvernement du Québec. Retrieved September 7, 2026, from https://www.revenuquebec.ca/en/businesses/consumption-taxes/gsthst-and-qst/registering-for-the-gst-and-qst/

Skok, D. (2013). SaaS metrics 2.0: A guide to measuring and improving what matters. For Entrepreneurs. Retrieved September 7, 2026, from https://www.forentrepreneurs.com/saas-metrics-2/

Van Westendorp, P. H. (1976). NSS-Price Sensitivity Meter (PSM): A new approach to study consumer perception of price. In Proceedings of the 29th ESOMAR Congress (pp. 139–167). ESOMAR.

Chapter 8 — Competitive Analysis and Positioning

"We have no competitors." Every experienced reader of a business plan has seen the sentence, and every one of them has read it the same way: the founder has not yet met the customer's alternatives. This chapter finds them — including the most common one, which is doing nothing — maps them on what the customer ranks, asks what advantage the venture can honestly claim, and chooses the position from which it will enter.

What You'll Be Able to Do

• Name the customer's real alternatives — direct competitors, indirect ones, substitutes, in-house solutions and doing nothing — and explain why the last is usually the first.
• Map the competitive landscape on an alternatives grid built from the customer's own rankings, using intelligence sources that are public and ethical.
• Apply the five industry forces lightly to a beachhead, and read what each force means for a venture that does not exist yet.
• Name an unfair advantage honestly, from the sources of advantage a small venture can actually build, and know how long each one lasts.
• Write the entry position of the new offer — the frame of reference, the category, the alternative and the difference — consistent with the value proposition of Chapter 4.
• Read barriers, regulation and timing as protections and as obstacles, and decide whether to be first or to follow fast.
• Run the competitive response test: what incumbents would do if the venture won its beachhead, and what the venture would do next.

Why It Matters

A venture does not fail because it has competitors; it fails because it did not know who they were, or because it could not say why a customer would choose it over them. Competitive analysis at the discovery stage is not the five-year strategy of Book A3; it is the shorter question of where to enter and what will protect the entry long enough to establish the model. For the founder, it is the antidote to two opposite illusions — that the field is empty, and that the incumbents are unbeatable. For a team, it is the argument that ends the debate about features: the customer does not compare us with the ideal; they compare us with what they do today. And for anyone who will ask for money, the alternatives grid and the unfair advantage are the two pages that the "no competitors" sentence would have cost them.

In This Chapter

• 8.1 Who You Really Compete With
• 8.2 Mapping the Competitive Landscape
• 8.3 Industry Forces for a Small Venture
• 8.4 Moats and Unfair Advantages
• 8.5 Positioning the New Offer
• 8.6 Barriers, Regulation and Timing
• 8.7 The Competitive Response Test

8.1 Who You Really Compete With

A competitor is whatever the customer would use instead. That definition is wider than the list of companies selling the same product, and narrower than the list of companies in the same industry, and the difference is what most founders get wrong in both directions.

8.1.1 Five kinds of alternative

The customer's alternatives come in five kinds, and the interviews of Chapter 3 have already listed them for the beachhead. Direct competitors sell the same kind of offer to the same segment — another receipt service, another compost collector. Indirect competitors solve the same problem with a different kind of offer — a full bookkeeping firm that includes the sorting in its fee, a general waste hauler that adds an organics bin. Substitutes serve the same job by other means — an app, a filing cabinet, a spouse who keeps the receipts. In-house solutions are the customer doing it themselves, with staff, a spreadsheet or a routine. And non-consumption — doing nothing, tolerating the problem, missing the deadline — is the alternative that most first customers are actually using, and the one that the value proposition of Chapter 4 must beat before any of the others.

DEFINITION Substitute A substitute is any offer, product, service or behavior that serves the same job for the customer by a different means, and to which the customer would turn if the venture's offer were unavailable or overpriced — including the customer's own labor and the option of doing nothing. Substitutes are the competitors that industry lists omit, and they set the ceiling on price: a customer will not pay more than the substitute costs them, in money and effort, plus the value of the difference.

8.1.2 Doing nothing is the first competitor

For a new offer, non-consumption is almost always the largest alternative in the segment, and it is the hardest to beat, because it is free, familiar and requires no decision. Chapter 2 treated non-consumption as a market gap — the people who have the job and do not buy the category; here it is read as the competitor: a founder who cannot say why a customer would stop tolerating the problem has no position, whatever the competitor grid says about the other companies. The cost of the status quo, in the customer's numbers, is the argument, and it was collected in the interviews.

KEY CONCEPT Your first competitor is doing nothing The industry has three suppliers and the founder has studied all three; the customer has met none of them and is using a shoebox. The competitive question of a new venture is not "how do we beat the incumbents?" but "why would this customer change what they do today?" — and the answer must beat inertia before it beats anyone else. A proposition that wins against the incumbents and loses against inertia sells to the few customers who were already shopping; a proposition that beats inertia opens the whole segment. The alternatives map of Section 8.2 therefore always has a "do nothing" column, rated honestly, and it is often the column the venture must beat most clearly.

8.1.3 The alternatives in the interviews

The list of alternatives is not compiled from a search engine; it is compiled from the answer to "what do you do about it today?" in twelve interviews. That answer names the real substitutes — the app three people abandoned, the assistant two people hired, the accountant who "handles it" for four — and the frequencies tell the founder which alternative is the incumbent in the segment. A search of the industry adds the direct competitors the customers did not mention, which is itself a finding: a competitor nobody in the segment named is not, for that segment, a competitor yet.

8.2 Mapping the Competitive Landscape

The landscape is mapped on a grid whose rows are what the customer ranks and whose columns are the customer's alternatives. The grid is the one instrument of this chapter that a founder cannot skip, because it is what the "only" test of Chapter 4, the positioning of Section 8.5 and the response test of Section 8.7 all read.

8.2.1 The alternatives grid

Figure 8.1 draws the grid for the receipt clinic. The rows are the attributes the customer ranked in the profile of Chapter 4 — the top pain relieved, the price, the effort to adopt, the risk of a penalty, the relationship — in the customer's order of importance. The columns are the alternatives, including doing nothing. Each cell rates, in the customer's judgment and words, how well the alternative serves the attribute. Read across a row to see who serves that attribute and who does not; read down the venture's column to see where it is strong, where it is merely adequate, and where it loses; and read the whole grid for the "only" claim — the attribute, high in the ranking, where the venture's cell is strong and every other cell is not.

Figure 8.1 — The alternatives map. Rows are the attributes the customer ranks — the top pain, price, effort to adopt, risk, relationship; columns are the real alternatives, including doing nothing; cells are the customers' own ratings. The clinic's "only" claim is read down its column: strong on the top pain, good on effort and relationship, mid-priced. Original figure; illustrative case.

METHOD The alternatives grid in one afternoon When to use it: after the profile and the value map exist, and before the positioning statement is final. Steps: (1) List the alternatives from the interviews' "what do you do today?" answers, with the count of people using each; add the direct competitors the customers did not name, marked as such. (2) Copy the ranked attributes from the customer profile — five to seven, in the customer's order. (3) Rate each alternative on each attribute from what customers said, not from what you think; where nobody said anything, mark the cell "unknown" rather than guessing. (4) Add the venture's column, rated as honestly — a founder who rates their own offer "strong" everywhere has drawn a wish. (5) Read the grid three ways: which alternative is the real incumbent in the segment; which attribute is the venture's "only"; which attribute the venture loses on, and whether it is high in the ranking. (6) Write the three findings under the grid, and date it. Deliverable: the grid with sources, the three findings, and the "only" sentence confirmed or revised. Pitfalls: rows chosen by the founder ("technology", "features") instead of the customer's ranking; a grid with no "do nothing" column; ratings from competitors' websites rather than from customers; and the most common — a venture column that is strong everywhere, which means the interviews were not consulted.

8.2.2 Sources of intelligence, and their rules

What a founder needs to know about competitors is mostly public, and it should be collected in public ways. Their own materials — websites, pricing pages, brochures, terms of service, case studies — say what they claim and to whom. Their customers' words — reviews, forum threads, the interviews' "what have you tried?" — say what they actually deliver. Public records — the enterprise register, corporate filings where available, trademark and patent databases, job postings, permits — say how big they are and where they are going. The market itself — trade shows, association meetings, suppliers' counters — says who is present and who is talked about. And the venture's own experience as a customer of the competitor, buying the product or the service openly, is legitimate and often the most instructive.

Three rules bound the collection. The venture does not misrepresent itself to obtain information — posing as a customer to extract a confidential quotation, or as a job applicant to learn a process, is the kind of deception that professional codes of competitive intelligence prohibit and that can expose the venture to legal claims (Strategic and Competitive Intelligence Professionals, n.d.). It does not use confidential information — a former employee's knowledge of a competitor's customer list or pricing is theirs to keep, not the venture's to use, and its use can breach the employee's obligations and the venture's. And it does not discuss prices, territories or customers with competitors: an agreement between competitors to fix prices, allocate markets or restrict supply is a criminal offence under the Competition Act, R.S.C. 1985, c. C-34 (s. 45), and a friendly conversation at an association meeting about "what we all charge" is where such agreements begin. The Québec and Canada Note at the end of this chapter returns to the rules.

8.2.3 Competitors as companies in the record

Every alternative worth tracking gets a record — a name, what it offers, its price if public, the segment it serves, its strengths and weaknesses in customers' words, and the date of the observation — because the grid will be redrawn quarterly and a founder who relies on memory redraws it from prejudice. La Machine's Companies page, described in the closing section, is one place to keep those records; a spreadsheet is another. Book B2 (Market Research and Competitive Intelligence) develops the monitoring into a continuous practice; at the discovery stage, one afternoon and a dated grid are enough.

8.3 Industry Forces for a Small Venture

Michael Porter's five forces describe the structure of an industry — the intensity of rivalry among existing competitors, the threat of new entrants, the threat of substitutes, the bargaining power of buyers and the bargaining power of suppliers — and explain why some industries are profitable for everyone in them and others for no one (Porter, 1979, 2008). Book A3 applies the framework in full. A venture in discovery applies it lightly, to the beachhead rather than to the industry, and asks of each force one question: what does it mean for a business that does not exist yet?

Table 8.1 states the five forces and, for each, what it means for a new venture, the question to ask about the beachhead, and what a founder can do about it. Read the table as a diagnostic rather than as a verdict: a segment with strong buyer power and weak rivalry is a different entry from one with weak buyers and a dominant incumbent, and the table says how.

Table 8.1 — The five forces, read for a beachhead

For a small venture, two of the forces deserve a special reading. Supplier power is often a software platform, a marketplace or a landlord, and it is the force founders underestimate: a model built on a platform's terms lives at the platform's pleasure, and the terms change. Buyer power is the large customer that every new venture wants and that turns the venture into a subcontractor when it arrives; the discovery quarter is the time to decide the maximum share of revenue any single customer may hold.

8.4 Moats and Unfair Advantages

An advantage is what lets the venture win a customer against the alternatives; an unfair advantage — a moat — is what keeps the venture winning after the alternatives have noticed and responded. The distinction matters because most first advantages are temporary, and the founder should know which kind they hold.

DEFINITION Moat (unfair advantage) A moat, or unfair advantage, is a feature of the venture's position that cannot be easily copied or bought by a competitor, and that lets the venture keep customers or margins after competitors respond. Ash Maurya's test in the Lean Canvas — something that cannot easily be copied or bought — is the practical one for a new venture (Maurya, 2012). A head start, a good idea, a lower price and hard work are advantages; they are not moats, because a competitor with resources can match each of them. A moat is what remains when they do.

8.4.1 The sources of durable advantage

Hamilton Helmer, surveying the strategy literature and the cases, identified seven "powers" that produce durable advantage: scale economies (unit costs fall with volume), network economies (the offer is worth more to each user as users increase), counter-positioning (a new model that incumbents cannot adopt without damaging their existing business), switching costs (customers lose value by leaving), branding (customers pay more for the same thing because of what the name means), a cornered resource (preferential access to an asset others cannot get) and process power (an embedded way of working that others cannot replicate quickly) (Helmer, 2016). Peter Thiel's shorter list for start-ups — proprietary technology, network effects, economies of scale and branding — overlaps it, with the same warning: a technology that is only marginally better is no advantage at all (Thiel & Masters, 2014).

8.4.2 The realistic moats of a small venture

Most of those powers belong to large firms or to platforms. A small venture rarely has scale, network effects are for two-sided models only, and a brand takes years. Table 8.2 lists the sources of advantage a small venture can actually build in its first two years. For each source it states what the advantage is, how a small firm builds it, how long it typically lasts against a determined competitor, and the test that tells whether the venture really has it. Read the "how long" column honestly: an advantage that lasts a year is worth having, and it is worth knowing that it is a year.

Table 8.2 — Sources of advantage a small venture can build

8.4.3 Naming the advantage honestly

The exercise ends in one sentence, written on the canvas and in the story of Chapter 6: "Our unfair advantage is [source], because [reason], and it should hold for [time] unless [what would break it]." Mariam's read: "Fifteen years of doing exactly this for exactly these owners, and the two accountants who refer them — which a new bookkeeper cannot reproduce in a year — and it should hold for two or three years unless a full firm decides to sell the sorting separately." A founder who cannot complete the sentence has a good idea and no moat, which is the normal condition of a first venture; the response is not despair but speed — win the beachhead before imitators arrive, and build the sources of Table 8.2 from the first customer.

WATCH OUT "We have no competitors" The sentence has three possible meanings, and none of them is good news. The founder has not asked customers what they do today, in which case the alternatives are unknown and the position is imaginary. The market has no competitors because nobody wants the offer, in which case the sentence is the timing signal of Chapter 2. Or competitors exist and the founder has defined them out of existence — "nobody does exactly this" — in which case the customer will make the comparison the founder refused to make. The consequence in all three cases is the same: a venture that meets its competitors for the first time in a sales conversation, unprepared. The alternatives grid, with its "do nothing" column, prevents all three.

8.5 Positioning the New Offer

Positioning is the place an offer occupies in the customer's mind relative to the alternatives — the idea that Al Ries and Jack Trout introduced and that Book B1 (Chapter 5) develops for a business with customers (Ries & Trout, 2001). For a new venture the question is narrower: from which position does the offer enter, and against which alternative does it define itself?

8.5.1 Three ways to enter

April Dunford describes three positioning styles that map onto a new venture's choices (Dunford, 2019). Head to head: enter an existing category and claim to be better than the leader on what the category's customers value — expensive, because the leader owns the category's attention, and reserved for a venture with a decisive advantage. A big fish in a small pond: enter an existing category but claim a sub-segment the leaders serve badly — the beachhead strategy of Chapter 5 expressed as a position, and the natural entry for most small ventures. Create a new game: define a new category that the offer leads by definition — the rarest and the most expensive, because the founder must teach the market that the category exists before selling into it. The alternatives grid says which one applies: a venture whose "only" attribute matters to a sub-segment is a big fish in a small pond; a venture whose "only" attribute defines a new job may be creating a game, and should first check that the job is real (Chapter 2).

8.5.2 The frame of reference

Every position needs a frame — the category the customer uses to understand what the offer is — and the frame decides which alternatives the customer compares it with and what price seems normal. "A bookkeeping service" frames the clinic against full firms and their $3,000 fees, which makes $720 look cheap and raises expectations of full service; "a receipt-sorting app" frames it against $15-a-month software, which makes $720 look absurd. Mariam's frame, "a quarterly bookkeeping session", was chosen because it existed in customers' minds (they knew what a session with a professional was), it carried the right price expectation, and it was not owned by anyone. The frame is tested in the repeat test of Chapter 4: if strangers say back a category the founder did not intend, the frame is wrong.

8.5.3 The entry position statement

The statement of Chapter 4 already contains the position: for whom, in which category, with which benefit, unlike which alternative, because of which difference. This chapter adds the two checks that competitive analysis makes possible. The alternative named in the statement is the real incumbent in the segment from the grid — the shoebox, not the full firm — and the difference is the "only" attribute confirmed by the grid, high in the customer's ranking. A statement that names an alternative nobody in the segment uses, or a difference nobody ranks, has been written from the founder's competitive anxieties rather than from the customer's map.

8.5.4 Repositioning

Positions move. A venture that enters as a big fish in a small pond may, after winning the pond, reposition toward the category's center (the accountants' firms, once the clinic is known) or toward a second pond (the next pin of Chapter 5). The discovery quarter fixes the entry position only, and records in the decision registry the alternatives considered and the signals that would prompt a move — the repositioning itself is a Book B1 exercise for a business with customers.

8.6 Barriers, Regulation and Timing

Some features of a market protect the venture that enters, others block it, and many do both depending on which side of them the venture stands. A founder reads them before entering.

8.6.1 Barriers that protect and barriers that block

Licenses, certifications and professional designations block the unqualified and protect the qualified; a founder who holds one has a moat, and a founder who needs one has a timeline and a cost (Book A4 lists the permits and licenses). Capital requirements — equipment, inventory, a facility — block entrants without financing and protect those who found it; the compost venture's truck is a barrier to the next compost venture. Distribution access — a shelf, a marketplace listing, a referral network — is a barrier that incumbents built and that a new venture must either buy or route around, which is why the beachhead criteria of Chapter 5 required a channel the venture already has. Customer habits and switching costs protect incumbents and are the reason "do nothing" is the first competitor. A founder lists the barriers of the beachhead in two columns — those the venture has crossed and those it has not — and the second column is a list of tasks or a reason to choose another beachhead.

8.6.2 Regulation as an obstacle and an opportunity

Regulation is a barrier in both directions. A rule that requires a permit, a certification or a compliance program is a cost and a delay for the venture, and the same rule is the reason customers need the venture: Norvale's restaurants bought compost collection because organic-waste rules were tightening, and the rule that made the sale possible also required the venture to be a compliant hauler. A founder checks, for the beachhead, what rules apply to the customers (which create demand) and what rules apply to the venture (which create obligations), and treats a coming rule change as the "why now" of Chapter 2 — with the caution that rules announced are not rules in force, and that a venture built for a rule that is delayed has a timing problem. Collection H gives the rules by domain.

8.6.3 First mover or fast follower

The founder who has found a gap wants to be first. The evidence says that being first is less valuable than it feels. Peter Golder and Gerard Tellis, studying the histories of fifty product categories, found that market pioneers failed at a high rate — nearly half — and that the long-term leaders of most categories were early followers who entered after the pioneer had proved the demand and then executed better (Golder & Tellis, 1993). For a small venture the lesson is practical: the advantage of being first is real only if the venture can convert it into one of the moats of Table 8.2 — references, switching costs, a cornered channel — before a follower arrives; otherwise the pioneer pays the cost of teaching the market and the follower collects the customers. A founder who is not first should not be discouraged; a founder who is first should not be complacent. Either way, the speed that matters is the speed of winning the beachhead, not the speed of launch.

8.7 The Competitive Response Test

The last step asks what happens after the venture wins. Competitors respond — or they do not — and the response is predictable enough to plan for.

8.7.1 What incumbents do

An incumbent that notices a new entrant winning a segment has five options. It can ignore the entrant, because the segment is too small to matter to it — the most common response to a small venture, and the one to hope for. It can match: cut its price, add the feature, launch a competing offer, which a large incumbent can do faster than a founder expects and a counter-positioned incumbent cannot do without damage. It can block: lock customers into contracts, sign exclusive terms with the channel the venture depends on, buy the supplier. It can acquire the entrant, which is a response some founders plan for (Book M1). Or it can lobby for rules that raise the entrant's costs. Table 8.3 lists the five responses with the signal that each is coming and the venture's own answer; read it as a plan written in advance rather than as a prediction.

Table 8.3 — Competitive responses and the venture's answers

8.7.2 Would they, could they, when

The test is three questions per alternative, answered in writing. Would they? Is the beachhead large enough, or the venture's entry visible enough, that the incumbent would notice and care within two years? Could they? Do the incumbent's economics and organization allow the response — a full firm can add a sorting session, but it cannot sell it at $720 without cannibalizing its $3,000 fee; a platform can add a feature in a month; a solo competitor can copy anything but cannot be in two places. When? How long would the response take, and would the venture have won the beachhead and built a moat by then? The answers go into the assumption map of Chapter 10 as risks, with the ones that would kill the model ranked first.

CASE IN POINT A meal-kit venture versus the grocery store Fictional example; names and figures are invented. Table Courte prepared weekly meal kits for families in one neighborhood — five dinners for four, ingredients pre-portioned, recipes in three languages — and its founders, two former restaurant cooks, wrote "no direct competitors in the area" in their first brief. The alternatives grid, drawn from fourteen interviews, corrected them. The real incumbent was the neighborhood grocery store's prepared-foods counter, which nine of fourteen families used two nights a week; a national meal-kit delivery service was the direct competitor, used by three and abandoned by two ("too much packaging, and I'm never home for the box"); and "cooking from what's in the fridge" — doing nothing — was the alternative on the other nights. The five forces said what the founders did not want to hear: the grocery chain had supplier power (it bought the same produce at a third of the price), buyer power was low (families spent $60 to $90 a week), substitutes were everywhere, and the barrier to entry was nil — a chain could launch kits in a month. The competitive response test asked what the chain would do if Table Courte won two hundred families. Would they notice? Two hundred families spending $75 a week is $780,000 a year, in one neighborhood — yes. Could they match? Yes, on price and reach, within a season. Would matching hurt them? No; the kits would sell more of their produce. The founders concluded that a head-to-head position on convenience was indefensible, and they repositioned before launch as a big fish in a small pond the chain could not enter without damage: kits built around a neighborhood chef's menu, with the chef's name on the box and a monthly dinner event, sold by subscription with pickup at a partner café — a relationship and a cornered resource (the chef, the café) that a chain cannot copy and would not want to. Two hundred families took eighteen months instead of six; the chain launched kits in month nine, as predicted; the venture kept 84% of its subscribers. The grid and the response test had not removed the competitor; they had chosen a position the competitor could not take.

QUÉBEC & CANADA NOTE Protecting what you build, and dealing with competitors lawfully (Canada and Québec, as of September 2026) Names and trademarks. A business name registered with the Registraire des entreprises (→ see Book A4) is not a trademark. A trademark — a word, design or combination that distinguishes the venture's goods or services — is registered with the Canadian Intellectual Property Office under the Trademarks Act, R.S.C. 1985, c. T-13 for a ten-year renewable term; before adopting a name, a founder searches the Canadian Trademarks Database and the enterprise register to avoid a demand letter after the brand has been built (Canadian Intellectual Property Office, 2026). Patents. An invention that is new, useful and non-obvious may be patented under the Patent Act, R.S.C. 1985, c. P-4; Canada allows a founder's own public disclosure within the twelve months before the filing date without destroying novelty (s. 28.2(1)(a), current to July 21, 2026) — a grace period that many other countries do not offer, so a founder who may want protection abroad should file before disclosing, or consult a patent agent first. Trade secrets — a method, a recipe, a customer list — are protected only by keeping them secret: confidentiality clauses in employment and supplier contracts, and access limited to those who need it; there is no registry. Copyright in written material, software, images and recordings arises automatically under the Copyright Act, R.S.C. 1985, c. C-42 and belongs to the author or the employer — which is why a designer's or a developer's contract must assign it to the venture. Book H5 (Intellectual Property) gives the full regime. Dealing with competitors. Agreements between competitors to fix prices, allocate customers or territories, or restrict supply are criminal offenses under the Competition Act, R.S.C. 1985, c. C-34 (s. 45), and the Competition Bureau prosecutes them; an association meeting is not a safe place to discuss what members charge. Comparative claims about a competitor must be true and not misleading (s. 74.01, Chapter 4). And obtaining a competitor's information by deception can engage civil liability as well as the professional codes of Section 8.2.2.

WHO DOES THIS Competitive analysis in the three situations In a solo venture the founder draws the grid from the interviews and reads it with an outsider who knows the industry — a supplier, a former colleague, an association member — because the founder's own ratings of competitors are the least reliable cells on the page. In a founding team, one member owns the grid and the records, and the team runs the response test as a debate in which one member argues the incumbent's side; the member who wrote the venture's column does not rate it. In an established business, the new offer's competitors are often the company's own existing products or partners, and the intrapreneur's first analysis is internal: which current offer the new one cannibalizes, which partner it competes with, and whether the company would match its own entrant; the answers decide whether the new line is launched inside the company or beside it. In all three, the grid is dated and redrawn quarterly, and the unfair advantage sentence is rewritten when it stops being true.

Costly Mistakes

1. Listing competitors from an industry search instead of from the interviews. The grid shows three companies the customers have never heard of and omits the shoebox, the app and the assistant they actually use. The consequence is a position defined against the wrong alternatives, and a first sales conversation in which the customer names an alternative the founder never considered.
2. Omitting the "do nothing" column. The venture beats every incumbent on paper and loses to inertia in every conversation, because the proposition never made the cost of the status quo visible. The consequence is selling only to the few customers who were already shopping.
3. Rating the venture's own column from ambition. Strong everywhere, weak nowhere, and the "only" claim rests on an attribute nobody ranks. The consequence is a differentiator that customers do not value and a founder surprised by indifference.
4. Confusing an advantage with a moat. A head start, a good idea and hard work are treated as protection, and the incumbent's matching response arrives in month nine with nothing built to withstand it. The consequence is a venture that taught the market and watched a follower collect the customers.
5. Entering head to head without a decisive advantage. Claiming to be better than the category leader on what the category's customers value, with the leader owning the attention and the channels. The consequence is the most expensive form of entry, paid for by a venture with the least money.
6. Building a brand on a name nobody searched. A demand letter after the website, the signage and the first hundred customers, and a rebrand at the worst possible moment. The consequence is money and reputation lost for want of an afternoon in the trademark and enterprise registers — and, in the other direction, discussing prices with competitors at an association meeting, which is a criminal offense, not a networking habit.

Key Takeaways

• A competitor is whatever the customer would use instead: direct and indirect competitors, substitutes, in-house solutions and doing nothing — and for a new offer, doing nothing is usually the largest alternative and the first to beat.
• The alternatives grid is built from the customer's rankings and the customer's answers, with a "do nothing" column and the venture rated as honestly as the rest; it yields the real incumbent, the "only" attribute and the attribute the venture loses on.
• Competitive intelligence is public and ethical: the competitors' own materials, their customers' words, public records and the market itself — never deception, never confidential information, and never a conversation with competitors about prices, customers or territories.
• The five forces are read for the beachhead, and two of them — supplier power in the form of a platform or a landlord, and buyer power in the form of one large customer — deserve a decision before the first sale.
• A moat is what cannot easily be copied or bought. A small venture's realistic moats are access, a documented method, switching costs, references within a niche, counter-positioning and a cornered resource; each has a lifespan, and the honest sentence names it.
• The entry position is a choice among three — head to head, a big fish in a small pond, or a new game — and for most small ventures it is the second; the frame of reference decides which alternatives and which price the customer has in mind.
• Being first is worth less than it feels: pioneers fail often and early followers lead most categories, unless the first mover converts its lead into a moat before the follower arrives; the speed that matters is the speed of winning the beachhead.
• The competitive response test — would they, could they, when — is written before launch, and the responses that would kill the model go to the assumption map as risks; the venture's answers to ignore, match, block, acquire and lobby are planned in advance.

Your Action Plan

The plan below takes about six hours over one week and produces the alternatives grid, the unfair advantage sentence and the entry position that the canvas, the story and the assumption map will carry. Steps 1 and 2 build the grid; step 3 reads the forces and the barriers; steps 4 and 5 name the advantage and the position; step 6 runs the response test and the name search.

In La Machine

La Machine keeps the competitors as records and finds them in the field; the grid itself is the founder's document.

• The Business Explorer (business/marketing/business-explorer) searches business directories and map sources by type and location — the same tool that listed the beachhead in Chapter 5 finds the direct competitors in the region, with their addresses and websites, and gathers them in the staging table with the source of each result.
• Companies (business/marketing/companies) hold one record per alternative worth tracking — the competitor's offer, price if public, segment, strengths and weaknesses in customers' words, and the date — with notes and linked contacts, so that the quarterly redraw of the grid starts from records rather than from memory. The industry and size fields, and the source and prospecting status carried over from the Explorer, describe each competitor as they describe a prospect.
• Links (/liens) keep the competitors' pricing pages and the trademark and enterprise registers one click away; Notes (/presse-papier) hold the findings under the grid and the response sheet, attached to the project.
• Arbitrages (/arbitrages) records the entry position chosen, the alternatives considered and the signals that would prompt a repositioning.

What La Machine does not do: it has no competitor-monitoring function, no alternatives grid and no register search — its alerts ring only at a chosen moment or on an event of the product, the grid is drawn on a whiteboard tool or a spreadsheet, and the trademark and enterprise searches are done on the registers' own sites. Book B2 (Market Research and Competitive Intelligence) turns the one-afternoon grid into continuous monitoring; Book A3 (Strategy and Competitive Positioning) develops the five forces and the moats into a strategy; Book H5 (Intellectual Property) covers trademarks, patents, copyright and trade secrets in full.

Self-Check

1. Name the five kinds of alternative a customer has, and explain why "doing nothing" is called the first competitor of a new offer.
2. What are the rows and the columns of the alternatives grid, where do the ratings come from, and what three things is the grid read for?
3. State the three rules that bound competitive intelligence, and the section of the Competition Act that makes one of them a criminal matter.
4. Distinguish an advantage from a moat, and name four sources of durable advantage that a small venture can realistically build.
5. What did Golder and Tellis find about market pioneers, and what does it imply for a founder who has found a gap?

Answers. (1) Direct competitors, indirect competitors, substitutes, in-house solutions and non-consumption; doing nothing is free, familiar and requires no decision, so it is usually the largest alternative in the segment and the one the proposition must beat before any other (Section 8.1). (2) Rows are the attributes the customer ranks, in the customer's order; columns are the real alternatives, including doing nothing, plus the venture; ratings come from customers' words in the interviews; the grid is read for the real incumbent in the segment, the "only" attribute, and the attribute the venture loses on (Section 8.2.1). (3) No misrepresentation to obtain information; no use of confidential information; no discussion of prices, customers or territories with competitors — the last is a criminal offense under Competition Act s. 45 (Section 8.2.2). (4) An advantage lets the venture win a customer against the alternatives; a moat keeps it winning after competitors respond, because it cannot easily be copied or bought; realistic sources for a small venture include access, a documented method, switching costs, references within a niche, counter-positioning and a cornered resource (Section 8.4). (5) Pioneers failed at a high rate — nearly half — and the long-term leaders of most categories were early followers who entered after demand was proved and executed better; a founder who is first must convert the lead into a moat — references, switching costs, a cornered channel — before a follower arrives, and the speed that matters is the speed of winning the beachhead (Section 8.6.3).

Chapter References

Competition Act, R.S.C. 1985, c. C-34, ss. 45, 74.01. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-34/

Copyright Act, R.S.C. 1985, c. C-42. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-42/

Patent Act, R.S.C. 1985, c. P-4, s. 28.2 (current to July 21, 2026). Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/P-4/

Trademarks Act, R.S.C. 1985, c. T-13. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/T-13/

Canadian Intellectual Property Office. (2026). Canadian Trademarks Database. Government of Canada. Retrieved September 7, 2026, from https://ised-isde.canada.ca/cipo/trademark-search/

Dunford, A. (2019). Obviously awesome: How to nail product positioning so customers get it, buy it, love it. Ambient Press.

Golder, P. N., & Tellis, G. J. (1993). Pioneer advantage: Marketing logic or marketing legend? Journal of Marketing Research, 30(2), 158–170.

Helmer, H. (2016). 7 powers: The foundations of business strategy. Helmer Associates.

Maurya, A. (2012). Running lean: Iterate from plan A to a plan that works (2nd ed.). O'Reilly Media.

Porter, M. E. (1979). How competitive forces shape strategy. Harvard Business Review, 57(2), 137–145.

Porter, M. E. (2008). The five competitive forces that shape strategy. Harvard Business Review, 86(1), 78–93.

Ries, A., & Trout, J. (2001). Positioning: The battle for your mind (20th anniversary ed.). McGraw-Hill.

Strategic and Competitive Intelligence Professionals. (n.d.). SCIP code of ethics for CI professionals. Retrieved September 7, 2026, from https://www.scip.org/

Thiel, P., & Masters, B. (2014). Zero to one: Notes on startups, or how to build the future. Crown Business.

Chapter 9 — Prototyping, MVPs and Pilot Programs

The model is drawn, the numbers close on paper and twelve people said the problem is real. The founder now wants to build. This chapter asks them to wait one more step — not to avoid building, but to build the smallest thing that can prove the model wrong, put it in front of real customers, and learn from what they do rather than from what they said.

What You'll Be Able to Do

• Explain why testing before building is cheaper than building before testing, using the build–measure–learn loop and the cost of learning.
• Choose among ten kinds of prototype — from a sketch to a paid pilot — the cheapest one that can change your mind about a given assumption.
• Define a minimum viable product correctly: minimum and viable, a full turn of the loop, and never half a product.
• Design a test with a test card: the hypothesis, the test, the metric, the threshold, the sample, the duration and the cost — and compute what the test is worth.
• Run a landing page or smoke test lawfully and honestly, with consent for the sign-ups and truthful claims on the page.
• Design and run a pilot program with selected customers, a written agreement and success criteria defined in advance.
• Run learning loops: what to change after each test, at what cadence, and when to stop.

Why It Matters

Building is the most expensive way to learn whether customers want something, and it is the way most founders choose, because building feels like progress and testing feels like doubt. The post-mortems of Chapter 1 record the result: products built for a market that was never asked. A test costs a day or a few hundred dollars and produces the behavioral evidence — sign-ups, payments, use, retention — that the interviews could not. For the founder, this chapter is where the discovery quarter earns its keep, because the tests either confirm the model at the top of the evidence ladder or refute it before the money is spent. For a team, tests replace opinions with numbers; for an established business, they replace the launch-and-hope habit with a gate that a new offer must pass.

In This Chapter

• 9.1 Test Before You Build
• 9.2 Types of Prototypes
• 9.3 The Minimum Viable Product
• 9.4 Designing a Test
• 9.5 Landing Page and Smoke Tests
• 9.6 Pilot Programs
• 9.7 Learning Loops and Iteration

9.1 Test Before You Build

Eric Ries described a start-up as an institution built to learn what customers want under extreme uncertainty, and the method he proposed — the lean start-up — runs on a loop: build the smallest thing that can test an idea, measure how customers respond, learn whether the idea holds, and go around again; the output of the loop is not a product but validated learning — knowledge about customers that has been tested against their behavior (Ries, 2011). The loop is the engine of this chapter, and the point of it is speed: the faster a founder goes around, the cheaper each lesson.

9.1.1 The cost of learning

Every assumption in the model can be tested in several ways, and the ways differ by a factor of a hundred in cost. Whether owners will pay $180 for a session can be learned by building a booking system, a brand and a service and launching them — months and thousands of dollars — or by asking seven interviewees for a deposit — an afternoon and nothing. Both produce the same lesson; one of them leaves the founder with the money to act on it. The principle that follows is the rule of this chapter: for every assumption, the founder chooses the cheapest test that could change their mind, and climbs to a more expensive test only when the cheap one has been passed and the next decision requires stronger evidence.

KEY CONCEPT The cheapest test that can change your mind A test is worth running only if one of its possible outcomes would change what the founder does next. A test whose result the founder would ignore — "if nobody signs up, we'll build it anyway" — is theatre. A test whose result would kill or redirect the model is the one to run first, and it should be the cheapest such test available. Figure 9.1 arranges the tests by cost and evidence: a founder starts at the bottom of the ladder and climbs only as far as the next decision requires. The question before every test is therefore not "what will this prove?" but "what would we do differently if it fails?" — and if the answer is "nothing", the test is not run.

DEFINITION Prototype A prototype is any artefact built to test one or more assumptions about the offer or the model before the offer exists in deliverable form — a sketch, a mock-up, a page, a video, a manual version of the service, a single working feature. A prototype is not a smaller product; it is a question posed in a form customers can answer with behavior. Its quality is measured by the evidence it produces per dollar and per day, not by its resemblance to the final offer.

9.1.2 The test ladder

Figure 9.1 draws five kinds of test as a staircase. On the first step, interviews and sketches establish that the problem is real — the evidence of Chapter 3, on the "says" rung. On the second, a landing page or smoke test measures interest at a price: do people who fit the profile leave an email, click a price, ask to be first? On the third, a concierge or Wizard of Oz test delivers the service by hand to a few paying customers and learns whether it works and what it takes. On the fourth, a single-feature minimum viable product learns whether customers use and keep a real, if narrow, offer. On the fifth, a paid pilot measures the outcome in real use, at a price, under a signed agreement. Each step costs more than the last and tests what the previous one could not; a founder who skips to the fifth pays for lessons the second would have taught.

Figure 9.1 — The test ladder. Interviews and sketches establish the problem; a landing page measures interest at a price; a concierge or Wizard of Oz test delivers the service by hand; a single-feature MVP measures use and retention; a paid pilot measures the outcome in real use under an agreement. Evidence strength rises with cost, and a founder climbs only as far as the decision requires. Original figure.

9.2 Types of Prototypes

The vocabulary of prototypes has grown around a few dozen practitioner techniques, which David Bland and Alexander Osterwalder cataloged and Alberto Savoia grouped under the name pretotyping — testing whether an idea is worth building before building it (Bland & Osterwalder, 2020; Savoia, 2019). Table 9.1 presents ten types that small ventures use most, in order of rising cost. For each it states what the type tests, which rung of the evidence ladder it reaches, its typical cost in hours and dollars, and the moment in the discovery quarter when it is used. Read the table as a menu keyed to assumptions: a founder picks the row that tests the assumption at hand at the lowest cost, not the row that looks most like the eventual product.

Table 9.1 — Ten types of prototype, from cheapest to costliest

9.2.1 Concierge and Wizard of Oz

Two rows of the table deserve explanation because they are the most useful and the least intuitive. In a concierge test, the founder delivers the service personally, by hand, to a handful of customers who know it is manual — a meal plan written each week in a messaging app, a receipt clinic run with paper envelopes, a recruiting service where the founder makes every call. Nothing is automated, nothing is built, and the customer pays; the founder learns what the service must do, which steps matter, what they cost in hours and where customers hesitate — in weeks, before a line of software. In a Wizard of Oz test, the customer sees what looks like a working product — a form, an app, a page — and the work behind it is done by a person; the customer's behavior toward the "product" is measured without the product existing. The two differ in what the customer knows: the concierge is openly manual, the Wizard is not, and Section 9.5's rule on honesty applies to both — the Wizard's customer must not be misled about what they are buying, only about how it is made.

9.2.2 Matching the prototype to the assumption

The choice of prototype follows the assumption, and Chapter 10 ranks the assumptions. A desirability assumption — "owners above $300,000 will pay $180 for a session" — is tested by a page with a price and a deposit, not by a mock-up. A feasibility assumption — "a session can be run in two hours by a bookkeeper who is not Mariam" — is tested by a concierge run by someone else. A viability assumption — "80% renew after the first year" — cannot be tested by any prototype in a quarter; it is tested by a pilot that runs long enough, or by the leading indicators of renewal (satisfaction, use, the "very disappointed" question of Chapter 4). A founder who maps each assumption to the prototype that can actually test it avoids the common error of building a beautiful mock-up to test a question about price.

9.3 The Minimum Viable Product

No term in this book is more used and less understood than the minimum viable product, and the misunderstanding costs ventures their first year.

DEFINITION Minimum viable product A minimum viable product (MVP) is the version of a new offer that allows a full turn of the build–measure–learn loop with the least effort and the least development time (Ries, 2011). It is minimum in that it does only what is needed to test the riskiest assumptions about the offer, and viable in that it actually delivers the core outcome to real customers, who use it and can pay for it. An MVP is not a prototype in the sense of Table 9.1's first rows — it is a real offer, narrow and rough — and it is not a smaller version of the eventual product with every feature reduced; it is the one feature that the top-ranked pain requires, done well enough to be used.

9.3.1 Minimum and viable

The two words pull against each other, and the tension is the point. Founders who hear minimum ship something so incomplete that it tests nothing — a form that goes nowhere, a page with no service behind it — and learn only that people do not use broken things. Founders who hear viable build the whole product and call it an MVP. The resolution is the value map of Chapter 4: the MVP delivers the top-ranked pain reliever completely and nothing else. A receipt clinic's MVP is one two-hour session with envelopes and a bookkeeper — no app, no portal, no year-end summary — delivered to ten customers who pay for it. A meal-planning service's MVP is a weekly plan and a grocery list, sent as a message, for the families who asked — no recipes, no app, no swaps. What is left out is not lost; it is the list of features whose demand the MVP will measure.

9.3.2 Not half a product

A widely used illustration contrasts two ways of building a car incrementally: delivering a wheel, then a chassis, then a body — useless until the last step — against delivering a skateboard, then a scooter, then a bicycle, then a car — each usable, each teaching something about what the customer wants from transport. The first is half a product; the second is a sequence of minimum viable products. The lesson for a founder is that the MVP must let the customer complete the job, however roughly, because a job half done produces no behavior to measure. If the job cannot be completed with the resources at hand, the answer is a concierge test — the founder completing it by hand — not a half-built tool.

9.3.3 The MVP for a service, a product and a platform

For a service, the MVP is almost always the founder delivering the core outcome personally, at a price, with the minimum of tools — the concierge test made into an offer. For a physical product, it is a small batch — hand-assembled, 3D-printed, bought and re-labeled — of the one variant the top pain requires, sold at a price, before tooling or inventory. For software, it is one workflow that solves the top pain end to end, often assembled from no-code tools rather than written, with everything else done manually behind it (the Wizard of Oz). For a platform, it is one side served by hand — the founder finding the bookkeepers, matching them to owners by email — until both sides show up on their own. In every case the MVP is measured on the same things: do customers complete the job, do they come back, do they pay, and what do they ask for next.

WATCH OUT Vanity metrics An MVP produces numbers, and the wrong numbers are the most comforting. Page views, downloads, sign-ups, followers, "users" who tried it once — each rises with effort and says nothing about whether the model works. Ries called them vanity metrics and contrasted them with actionable metrics — the ones that show cause and effect and change a decision (Ries, 2011). For a small venture the actionable numbers are few: the share of people who fit the profile and complete the job; the share who come back a second and a third time; the share who pay at the price hypothesized; and the reason, in their words, when they stop. A test card that names a vanity metric as its measure will pass whatever the truth is; the threshold of Section 9.4 is written on an actionable one.

9.4 Designing a Test

A test is designed before it is run, on one page, so that its result can be read afterwards without argument. The instrument is the test card, adapted from the format that Bland and Osterwalder use for every experiment (Bland & Osterwalder, 2020).

METHOD The test card When to use it: before every test on the ladder, from a landing page to a pilot. Steps — the seven fields: (1) The hypothesis, in the form of Chapter 1 — who, what, how many, threshold; Savoia's version is useful: "at least X% of Y will Z" — at least 8% of owners above $300,000 who visit the page will leave an email and a preferred date. (2) The test: what will be built and shown to whom, through which channel — a one-page description of the clinic with a price and a booking form, sent to the association's 1,200 members by its newsletter. (3) The metric: the actionable number that will be counted — sign-ups with a date, not visits. (4) The threshold: the number at which the hypothesis is validated, set before the test and written down — 8% of visitors; below 3%, invalidated; between, inconclusive. (5) The sample: how many observations are needed for the metric to mean anything — at least 100 fitting visitors for a rate, five users for a usability session (Nielsen, 2000). (6) The duration: when the test ends, whatever the count — two weeks. (7) The cost: hours and dollars, so that the value of the test can be judged. Deliverable: one dated card per test, filed in the validation log, with the result and the decision added when the test ends. Pitfalls: a hypothesis without a number; a threshold set after seeing the result; a metric that is a vanity metric; a sample so small that one person decides the outcome; a test with no end date, which runs until it says what the founder wants.

9.4.1 Thresholds before results

The threshold is the discipline of the whole method. Set before the test, it turns a result into a decision; set after, it turns a decision into a rationalization — "3% is actually pretty good for a first page". The threshold comes from the model: if the weekly number of Chapter 5 requires eight contacts a week from the association, and the newsletter reaches 1,200 people quarterly, the page must convert about 8% of fitting visitors for the channel to work; below that, the channel or the proposition must change. A founder who cannot derive the threshold from the model has not understood what the test is for.

9.4.2 Samples and how many is enough

Two rules of thumb cover most tests. For a rate — sign-ups per visitor, purchases per sign-up — a hundred observations give a margin of about ten points, as the arithmetic of Chapter 3 showed; fewer than thirty makes the result an anecdote. For usability — can people find their way through the offer — Jakob Nielsen's finding holds: five users in a session reveal most of the problems, and further users mostly repeat them; the remedy is to fix and test again with five more (Nielsen, 2000). For a concierge or a pilot, the sample is the number of customers the founder can serve by hand — five to ten — and the evidence is deep rather than wide: what each of them did, week by week.

DO THE MATH What a test is worth Illustrative figures. Decision analysis gives a way to value information before buying it: the expected value of a test is roughly the probability that it changes the decision, times the cost of the mistake it would prevent, minus the cost of the test (Howard, 1966). Mariam believes, with about 70% confidence, that owners will pay $180 a session; if she is wrong, she will have spent about $4,500 on a booking system, kits and a launch before finding out, and lost a quarter. A two-week page test with a price and a deposit costs her twelve hours and $150. The chance that it changes her mind is the 30% she is uncertain by — call it 25%, since even a failed test might be inconclusive. Value ≈ 0.25 × $4,500 − ($150 + 12 hours at $60) = $1,125 − $870 = about $255, plus the quarter saved, which the arithmetic undervalues. The same test for a decision worth $40,000 — a lease, a hire — is worth about $9,000. The rule the arithmetic teaches is simple: the more expensive the next commitment, and the less certain the founder, the more a cheap test is worth; and a test that could not change the decision (probability zero) is worth nothing at any price.

9.5 Landing Page and Smoke Tests

The landing page is the workhorse of discovery: a single page that describes the offer as if it existed, at a price, with one call to action, put in front of people who fit the profile. It is cheap, fast, and it produces the first behavioral evidence a venture has — provided it is run honestly.

9.5.1 How to run one

The page carries the value proposition statement of Chapter 4 in its tested form, the top three pain relievers in the customer's words, the price or a price range, one piece of proof that is true (the founder's credentials, "in twelve interviews, nine…"), the stage stated plainly ("first sessions in January"), and one call to action — join the waiting list, book a discovery call, pre-order with a refundable deposit. Traffic comes from the beachhead's own channels — the association's newsletter, the referring accountants' introductions, the founder's network, a small paid campaign targeted at the segment if the channel exists — and never from a general audience, because the rate that matters is the rate among people who fit. The metric is the actionable one from the test card: sign-ups with a preferred date, deposits, calls booked. A second page with a different price, shown to a different half of the traffic, turns the test into a price test.

9.5.2 Reading the result

The result is read against the threshold. A sign-up rate among fitting visitors in the range that practitioners commonly report for targeted, single-offer pages — roughly 5% to 15%, a range that varies widely by offer and by traffic and is quoted here only to set expectations — is a signal that the proposition lands; a rate near zero with fitting traffic is a signal about the proposition, the price or the frame; a rate near zero with untargeted traffic says nothing. The page also produces words: the questions people ask in the form's free field, the objections in replies, the "when can I have it?" — all of which go into the log beside the numbers.

THE RULE Honesty and consent on a test page (Canada and Québec, as of September 2026) Three rules apply to a landing page from the moment it is public. Truthful claims. A page that describes an offer as existing when it does not, promises results it has not produced, or states numbers it cannot substantiate makes a representation that is false or misleading in a material respect, prohibited under the Competition Act, R.S.C. 1985, c. C-34 (s. 74.01); state the stage, and keep every claim in the claims file of Chapter 4. Consent for the emails. A sign-up on a waiting list is a request for future messages, and those messages are commercial electronic messages under Canada's anti-spam legislation, which requires express consent obtained with a clear description of the purpose and the identity of the sender, and a working unsubscribe mechanism in every message (CASL ss. 6, 10); the sign-up form must therefore say who will write, about what, and how to stop, and the consent to marketing must be a separate, unchecked choice from the delivery of whatever the page promises — the unbundled consent that Québec's Act respecting the protection of personal information in the private sector, CQLR c P-39.1 requires (ss. 8, 14). Consumer pre-orders. If the page takes deposits from consumers, the distance-contract rules of the Consumer Protection Act, CQLR c P-40.1 summarized in Chapter 3 apply — disclosure before the contract, payment by a means that preserves the chargeback right, a stated delivery date, refunds within 15 days of a cancellation (ss. 54.3–54.13). And in Québec, a commercial web page must be available in French under the Charter of the French language, CQLR c C-11 (Book B1, Chapter 10; Book H4). None of these rules prevents a test; each one shapes the page, and a page built to them is more convincing to early adopters than one built to look finished.

9.5.3 Fake doors and the limit of deception

A fake door — a button for a feature that does not exist, which records the click and then explains that the feature is coming — is a legitimate test inside an existing product, where the user loses nothing but a click and is told the truth immediately. On a public page for a new venture, the same technique shades into the misleading representation the rule above prohibits, and it costs the trust of the very people the venture needs. The safe form is transparency: "coming in January — tell us if you want to be first" tests the same interest without pretending the door leads anywhere yet.

9.6 Pilot Programs

At the top of the ladder stands the pilot: a real delivery of the offer to a few selected customers, for a defined period, under an agreement, with success criteria written before it starts. A pilot is what a lender, an investor or a large customer means by "proof", and it is the test that measures the outcome rather than the interest.

DEFINITION Pilot program A pilot program is a bounded, real delivery of the offer to a small number of selected customers, for a stated period and usually at a price, under a written agreement that sets what will be delivered, what will be measured, what counts as success, and what happens at the end — conversion to a regular customer, extension, or a stop. A pilot tests the outcome in real use, the venture's ability to deliver, and the customer's willingness to pay for the result; it produces the references and the measured results that the venture's first claims will rest on.

9.6.1 Selecting pilot customers

The best pilot customers fit the ideal customer profile exactly, have the problem now, will give feedback honestly and often, are willing to be named as a reference if the pilot succeeds, and are not the largest or most prestigious prospect available — a large customer's demands bend the offer toward that one customer, and its silence when the pilot fails is expensive. Three to ten pilots are typical for a service; more than the founder can attend to personally is too many. Interviewees who asked "when can I have it?" and paid a deposit are the natural first pilots; the referring channel of the beachhead supplies the rest.

9.6.2 Paid or free

A paid pilot is better evidence than a free one, because payment is the top rung of the ladder and because a customer who pays behaves like a customer — they use the offer, complain, and expect the outcome. A discounted pilot with reference rights in exchange is the usual compromise: the customer pays a reduced price and agrees to be quoted, to complete the measurement and to give a testimonial if the criteria are met. A free pilot is defensible only when the offer's cost to the customer is mostly their own time and the venture's need is the measurement; even then, a token payment changes the customer's behavior more than its amount suggests.

9.6.3 The pilot agreement

Table 9.2 is the checklist of what a pilot agreement covers. It is written on one or two pages, in plain language, and it protects both parties less by its legal force than by the conversation it obliges before the pilot starts: what exactly will be delivered, what will be measured, and what "success" means. Read the table as the agenda of that conversation; Book H1 (Contracts and Commercial Agreements) gives the clauses for a fuller contract, and Book E5 (Proposals, Quotes, Business Cases and Contracts) the proposal that precedes it.

Table 9.2 — The pilot agreement checklist

9.6.4 Running and ending the pilot

A pilot is run like a customer engagement and measured like a test. The founder delivers personally or watches closely, meets the customer at the agreed cadence, logs everything — use, incidents, questions, hours spent, costs — and measures the outcome metric against its baseline. At the end, the customer is asked the question of Chapter 4 — how would you feel if you could no longer use this? — and the pilot converts, extends or stops according to the agreement. A pilot that meets its criteria produces a reference, a measured result for the claims file and a customer at the regular price; a pilot that fails produces the most valuable finding of the quarter, and the founder treats it as such.

9.7 Learning Loops and Iteration

Tests produce results; iteration turns results into a better model. The loop has a cadence, a rule about what to change, and a rule about when to stop.

9.7.1 Cadence

Ries's loop turns as fast as the cheapest test allows. For a landing page, a turn is two weeks — a page, traffic, a result, a revised page. For a concierge test, a turn is one delivery cycle — a week for a meal plan, a quarter for a receipt clinic, which is why the clinic's concierge test starts with a single session rather than a year. For a pilot, a turn is the pilot's duration, which is why pilots are short and few. Jake Knapp and his colleagues compressed the loop for a prototype into a five-day sprint — map the problem on Monday, sketch solutions on Tuesday, decide on Wednesday, build a realistic prototype on Thursday, test it with five customers on Friday — a format that a founding team can run once a month (Knapp, Zeratsky, & Kowitz, 2016). Chapter 11 places the loop in the two-week discovery sprint.

9.7.2 What to change

Two rules govern iteration. One variable at a time where the test allows it: a page whose price, headline and call to action all changed between two runs has taught nothing about any of them. And the riskiest assumption first: the next test addresses the assumption whose failure would most damage the model and about which the evidence is weakest — the assumption map of Chapter 10 ranks them — rather than the assumption the founder finds most interesting. A learning card, the companion of the test card, records each turn: what was tested, what was observed, what was learned, and what will be done — and the "what will be done" is the next test card.

9.7.3 When to stop

Iteration is not infinite. A test has an end date; a series of tests has a stop rule, written before the series begins: the number of turns, the money and the time after which, if the threshold has not been reached, the founder declares the assumption invalidated and moves to the decision of Chapter 10 — persevere on a revised model, pivot or stop. Without the rule, a founder iterates a landing page for six months, each version "almost there", and the discovery quarter becomes a discovery year. The kill number of Chapter 7 — the price, the rate, the count below which the model does not work — is the stop rule's substance.

CASE IN POINT A concierge MVP for a meal-planning service Fictional example; names and figures are invented. Semaine Prête was to be an app that planned a family's weekly dinners around their preferences and their grocery store's flyer. Its founder, a dietitian, had twelve interviews with working parents that confirmed the pain — "Sunday night, no plan, and the same four meals" — and a landing page with a $15-a-week price that converted 9% of the parents reached through two daycare newsletters. Before building anything, she ran a concierge MVP: eight families, $15 a week paid by e-transfer, and every Sunday afternoon she wrote each family's plan by hand and sent it as a message — five dinners, a grocery list, nothing else. The test card said: at least six of eight will pay for a fourth week, and the average family will cook at least four of the five dinners, measured by a Sunday-evening question. Six weeks taught her more than a year of building would have. Seven of eight paid for the fourth week. The recipes she had planned to write were never asked for; the grocery list, ordered by aisle, was mentioned by every family as the thing they used. Sixty percent of the messages she received were swap requests — "we're out on Thursday", "no fish this week" — which the app had not been designed to handle and which took her, by hand, ten minutes per family per week. And two families asked whether she could order the groceries, which she had not imagined. The single-feature MVP that followed was not the recipe app; it was a plan-and-list generator with a swap button, assembled from no-code tools in three weeks, with the ordering left for a later pin. The concierge had cost her six Sunday afternoons and $0; the app she had almost built had been quoted at $34,000.

WHO DOES THIS Testing in the three situations In a solo venture the founder runs every test and, in a concierge, is the product, which is exactly the point: the founder learns what the service takes before deciding what to automate or hire; the discipline is the test card written before, and the stop rule, because nobody else will enforce either. In a founding team, one member owns each test card, the team reviews results together against the thresholds, and the member who built the prototype does not read its result alone; the five-day sprint is the team's natural rhythm. In an established business, pilots are the normal instrument — the company has customers to pilot with — and the risks are different: the sales force wants to sell the pilot as a product, the customer expects the company's full service, and the pilot's cost is hidden in existing salaries; the intrapreneur writes the agreement, the criteria and the exit as if the pilot were external, and reports the results at the gate as numbers against thresholds. In all three, the actionable metrics are named before the test and the vanity metrics are not reported.

Costly Mistakes

1. Building before testing. The booking system, the brand and the inventory are bought for a proposition that a $150 page test would have refuted in two weeks. The consequence is the year and the tens of thousands of dollars of Chapter 1's Do the Math box, spent learning what the second step of the ladder teaches.
2. A test that cannot change the decision. The founder runs a page "to see", with no threshold and a plan to build regardless. The consequence is theatre: a number that means nothing, reported as validation.
3. Half a product called an MVP. A form that goes nowhere, a service that cannot complete the job, and the conclusion that "people don't use it". The consequence is a false negative on a real opportunity, and a founder who abandons a model that a concierge test would have confirmed.
4. Thresholds set after the result. "Three percent is pretty good for a first page." The consequence is a discovery quarter that validates everything and a model built on rationalizations.
5. A dishonest test page. A product described as existing, results it has not produced, a fake door on a public page, a waiting list with no consent language. The consequence is exposure under the Competition Act and the anti-spam legislation, and — more immediately — the loss of the early adopters' trust, which a new venture cannot buy back.
6. A pilot without an agreement or criteria. An open-ended free service to a large customer, measured on nothing, ending in "it was interesting". The consequence is months of delivery, no reference, no measured result, and a customer who expects the free service to continue.

Key Takeaways

• Building is the most expensive way to learn, and the loop — build the smallest thing, measure behavior, learn, repeat — is the engine of the discovery quarter; its output is validated learning, not a product.
• Choose the cheapest test that can change your mind, and climb the ladder — interviews, landing page, concierge or Wizard of Oz, single-feature MVP, paid pilot — only as far as the next decision requires.
• Match the prototype to the assumption: a desirability assumption is tested by a page with a price and a deposit, a feasibility assumption by a concierge run by someone else, a viability assumption by a pilot that runs long enough.
• A minimum viable product is minimum and viable: the one feature the top-ranked pain requires, done well enough that real customers complete the job and can pay — never half a product, and for a service usually the founder delivering by hand.
• Every test has a card written before it runs: hypothesis with a number, test, actionable metric, threshold, sample, duration, cost — and a threshold set afterwards is a rationalization.
• A landing page is run on the beachhead's channels, honestly: the stage stated, every claim in the file, express and unbundled consent for the waiting list, and the consumer rules for deposits.
• A pilot is a real delivery under an agreement: selected customers who fit the profile, a price or a discount with reference rights, success criteria and an exit decided before the start.
• Iteration has a cadence, a rule and a stop: one variable at a time, the riskiest assumption first, and a stop rule written before the series — the kill number of Chapter 7 — that ends the loop and starts the decision of Chapter 10.

Your Action Plan

The plan below runs over about five weeks of the six-hour slot and produces the behavioral evidence that Chapter 10's decision will rest on. Steps 1 and 2 design the first tests; steps 3 and 4 run the page and the concierge; step 5 climbs to a pilot where the decision requires it; step 6 closes each loop with a learning card.

In La Machine

La Machine serves the test pages, captures the sign-ups — with a documented consent when the capture comes through a lead magnet's consent box — and holds the test plan; the measurement and the pilots are the founder's work, recorded in the platform. The pages below are described as read on September 15, 2026; Book N5 gives each of them in full.

• Landing pages (business/marketing/landing-pages) are the test pages of Section 9.5: one page per offer or price, previewed with the Preview button of its editor, and served publicly at /lp/:slug only when it is published and active — the public address refuses drafts. Publishing is a status in the platform, not a lock, so the human approval of each publication is the business's written rule, checked against the claims file. Each submission creates or updates a contact and records a response, gathered at business/marketing/landing-pages/responses, and the visit count comes from a journal of successful public views, so that the sign-up rate of the test card can be computed from visits and responses.
• Lead magnets (business/marketing/lead-magnets) and forms (business/marketing/forms) capture the waiting-list sign-ups and the pre-order requests. Only the lead magnet's own consent box — unticked, optional, separate from the delivery of what the page promises — documents a consent at capture, and its label, which names who will write and about what, becomes the proof sentence; a checkbox added to a form records an answer, not a consent. A waiting list that will receive messages is therefore captured through a lead magnet, which gives the consent that the rule of Section 9.5 requires.
• Sequences (business/marketing/sequences) are the one path by which the module sends. A contact enrolled in an active sequence is advanced by a background engine, and each message passes the dispatcher's checks — the master exclusion list, the sender's identification and the consent flag — before the email transport sends it with an instruction to reply to unsubscribe; a message that fails a check is stopped, not sent. Marking a sequence active is therefore a decision with effect when its trigger is an event, and only email is delivered. The email queue page (business/marketing/email-queue) is a manual fallback over a table the sending path does not use, and its Mark as sent action sends nothing.
• Projects and tasks (/projets, /taches) hold the test plan, each test card as a task with its criterion and its end date, which is set when the task is loaded because no page edits it, so that the cadence the agreement of Table 9.2 sets is scheduled and tracked rather than remembered.
• The website builder (business/website/builder) builds a fuller page when a single landing page is not enough.

What La Machine does not do: it has no split-testing function — two prices are tested with two pages and two audiences, and the comparison is made by the founder — no analytics beyond visits and responses — opens and clicks are not measured — and no payment collection for deposits on a test page at this stage, which is done with a payment tool of Chapter 13 within the consumer rules. The pilots are delivered and measured by the founder; the learning cards live in the validation log. Book B9 (Lead Generation and Lead Magnets) develops the pages and the magnets, Book C2 (Websites, Landing Pages and Conversion) develops conversion, and Book N5 (Marketing Automation in La Machine) covers the pages, forms, sequences and the sending path of the platform page by page, as of September 15, 2026.

Self-Check

1. State the principle of "the cheapest test that can change your mind", and say what question a founder asks before running any test.
2. Name the five steps of the test ladder and the rung of evidence each one reaches.
3. Define a minimum viable product and explain why "half a product" is not one.
4. List the seven fields of a test card and explain why the threshold must be set before the test.
5. Name three legal rules that apply to a public landing page for a new venture in Québec, and the item of a pilot agreement that a pilot most often lacks.

Answers. (1) For every assumption, run the cheapest test whose possible outcomes include one that would change what you do next, and climb to a costlier test only when the next decision requires stronger evidence; the question is "what would we do differently if this fails?" — if nothing, the test is not run (Section 9.1.1). (2) Interviews and sketches — says; a landing page or smoke test — does; a concierge or Wizard of Oz test — does and pays; a single-feature MVP — does and pays, with use and retention; a paid pilot — pays, with the outcome measured under an agreement (Section 9.1.2). (3) The version of a new offer that allows a full turn of the build–measure–learn loop with the least effort — minimum in doing only what tests the riskiest assumptions, viable in delivering the core outcome to real customers who can pay; half a product does not let the customer complete the job, so it produces no behavior to measure and a false negative (Section 9.3). (4) Hypothesis with a number, test, metric, threshold, sample, duration, cost; set before, the threshold turns a result into a decision; set after, it turns a decision into a rationalization (Section 9.4). (5) Truthful claims under Competition Act s. 74.01; express, unbundled consent and identification and unsubscribe for waiting-list messages under CASL ss. 6, 10 and P-39.1 ss. 8, 14; the distance-contract rules of the Consumer Protection Act for consumer deposits, and French under the Charter of the French language; the item most often lacking is the success criteria with baseline, target and measurer agreed before the start (Sections 9.5–9.6).

Chapter References

An Act to promote the efficiency and adaptability of the Canadian economy by regulating certain activities that discourage reliance on electronic means of carrying out commercial activities, S.C. 2010, c. 23 (Canada's Anti-Spam Legislation), ss. 6, 10. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/E-1.6/

Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25), ss. 8, 14. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1

Charter of the French language, CQLR c C-11. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/C-11

Competition Act, R.S.C. 1985, c. C-34, s. 74.01. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-34/

Consumer Protection Act, CQLR c P-40.1, ss. 54.1–54.16. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-40.1

Bland, D. J., & Osterwalder, A. (2020). Testing business ideas: A field guide for rapid experimentation. Wiley.

Howard, R. A. (1966). Information value theory. IEEE Transactions on Systems Science and Cybernetics, 2(1), 22–26.

Knapp, J., Zeratsky, J., & Kowitz, B. (2016). Sprint: How to solve big problems and test new ideas in just five days. Simon & Schuster.

Nielsen, J. (2000, March 18). Why you only need to test with 5 users. Nielsen Norman Group. Retrieved September 7, 2026, from https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/

Ries, E. (2011). The lean startup: How today's entrepreneurs use continuous innovation to create radically successful businesses. Crown Business.

Savoia, A. (2019). The right it: Why so many ideas fail and how to make sure yours succeed. HarperOne.

Chapter 10 — Risk, Assumptions and Pivots

Two tests have run. One passed, one failed, and the founder is holding a page of results, a canvas with thirty notes and a deadline they set for themselves in Chapter 1. This chapter is where the discovery quarter becomes a decision: which assumptions were at stake, which ones the evidence has settled, and whether the venture should persevere on this model, pivot to another, or stop — and how to make that decision in a way the founder will still respect a year later.

What You'll Be Able to Do

• Extract the assumptions hidden in a business model and sort them into the four families of risk: desirability, feasibility, viability and adaptability.
• Rank assumptions on an assumption map — importance against evidence — and pick the three to five that must be tested first.
• Run the riskiest assumption first, sequence tests by dependency, and use discovery-driven planning to know which milestone each assumption blocks.
• Read test results honestly — validated, invalidated, inconclusive — weighing the strength of the evidence and resisting the pull of confirmation.
• Decide to persevere, pivot or stop on written criteria, and recognize the ten kinds of pivot and the signals that call for each.
• Keep a risk register for the risks that are not assumptions — legal, financial, operational, personal — and know the founder's own exposure.
• Decide with discipline: decision records, a pre-mortem before each commitment, and a registry of decisions that survives the founder's memory.

Why It Matters

Every venture is a bet on a set of beliefs, and most ventures fail on a belief nobody wrote down. The founder who can list the beliefs, rank them by how much damage their falsity would do and how little is known about them, and test the worst first, converts a bet into a sequence of small, cheap questions. The founder who cannot spends the same money on the wrong questions and learns the fatal answer last. For the founder, this chapter is the difference between a pivot at week ten and a failure at month fourteen. For a team, it is the shared discipline that stops the persevere-or-pivot argument from being won by whoever cares most. For an established business, it is the gate that a new offer must pass on evidence rather than on sponsorship.

In This Chapter

• 10.1 Every Model Is a Set of Assumptions
• 10.2 Ranking Assumptions
• 10.3 The Riskiest Assumption Test
• 10.4 Reading Results
• 10.5 Persevere, Pivot or Stop
• 10.6 Risk Beyond Assumptions
• 10.7 Deciding with Discipline

10.1 Every Model Is a Set of Assumptions

Chapter 1 said that a model is a set of hypotheses until evidence says otherwise; Chapter 6 marked every note on the canvas as evidence or hypothesis; Chapter 7 wrote the assumptions table with its three numbers that must be true; Chapter 8 added the competitive responses that would kill the model. This section gathers all of it into one list and gives it structure.

DEFINITION Assumption An assumption is a belief on which the business model depends and which has not yet been established by evidence — about the customer, the offer, the channel, the numbers, the venture's ability to deliver, or the environment. Every assumption can be written as a hypothesis in the form of Chapter 1 (who, what, how many, threshold), and a hypothesis is the testable form of an assumption. The distinction matters in practice: founders hold many assumptions they have never noticed, and the first task of this chapter is to notice them.

10.1.1 Extracting assumptions from the canvas

The method is mechanical, which is its virtue. For each note on the canvas, the founder writes the sentence that must be true for the note to hold: "owners above $300,000 with staff exist in the region in sufficient number"; "they will pay $180 a session"; "the association will publish the offer to its members"; "a bookkeeper other than Mariam can run a session in two hours"; "80% will renew after the first year"; "no full firm will unbundle the sorting within two years". Thirty to fifty sentences is normal for a first canvas, and the length is not a problem — it is the inventory that the map of Section 10.2 will shorten.

10.1.2 Four families of risk

Bland and Osterwalder sort the assumptions of a model into four kinds of risk, which extend the three fits of Chapter 1 (Bland & Osterwalder, 2020). Desirability risk: the customers do not want it — the segment is wrong, the pain is mild, the proposition does not resonate, the channel does not reach them, the relationship is not what they expect. Feasibility risk: the venture cannot deliver it — the key resources, activities and partners are unavailable, unaffordable or slower than assumed. Viability risk: the numbers do not work — the price is too low, the costs too high, the acquisition too expensive, the renewal too weak, the runway too short. Adaptability risk: the environment does not allow it — a regulation, a competitor's response, a trend that reverses, a supplier's terms. Each assumption on the list is tagged with its family, and the tag says which chapter's instruments test it: Chapters 3 and 9 for desirability, Chapters 6 and 12 for feasibility, Chapter 7 for viability, Chapter 8 for adaptability.

10.1.3 Leap-of-faith assumptions

Ries singled out two assumptions that every new venture makes and that dominate all the others: the value hypothesis — that the offer delivers value to customers once they use it — and the growth hypothesis — that new customers will find the offer through a mechanism the venture can sustain (Ries, 2011). In the vocabulary of this book, the first is the desirability of the value proposition at the top of the evidence ladder, and the second is the channel and the weekly number of Chapter 5. They are the assumptions a founder is most tempted to take on faith, because testing them means facing the customer, and they are almost always in the upper-left corner of the map.

10.2 Ranking Assumptions

Fifty assumptions cannot all be tested in a quarter. The assumption map ranks them on two questions — how important is this assumption, and how much evidence do we have? — and points at the few that must be tested first.

10.2.1 Importance and evidence

Importance asks: if this assumption turned out to be false, would the model fail, or merely change? An assumption whose falsity kills the model — nobody will pay, the segment cannot be reached, the service cannot be delivered at the cost — scores 5; one whose falsity changes a detail — the year-end summary is not wanted — scores 1. Evidence asks: how much do we already know, and from which rung of the ladder? An assumption supported by payments from strangers scores 5; one supported by the founder's conviction scores 1. Figure 10.1 draws the map with example assumptions from the receipt-clinic case: the renewal rate and the weekly contacts are important and unsupported, and they are tested first; the price is important and now supported by deposits, and it is monitored; the year-end summary is unimportant and unknown, and it waits.

Figure 10.1 — The assumption map. The vertical axis is importance — would the model fail if the assumption were false? The horizontal axis is evidence — how much do we know, and from which rung? Important and unsupported assumptions are tested first; important and supported ones are monitored; the rest wait or are ignored. Assumptions move right as evidence arrives, or are struck out. Original figure, adapted from the concept described by Bland and Osterwalder (2020); illustrative placements from the fictional case.

METHOD The assumption map in sixty minutes When to use it: after the canvas is drawn (Chapter 6) and the numbers are stress-tested (Chapter 7), and again after every test. Steps: (1) Write every assumption from the canvas, the assumptions table and the competitive response test on its own note — one sentence each, in the testable form. (2) Score each on importance from 1 to 5: would the model fail if this were false? (3) Score each on evidence from 1 to 5, using the rung and the source: 1 for the founder's belief, 2 for what a friend said, 3 for what strangers said, 4 for what strangers did, 5 for what strangers paid. (4) Place the notes on the map; the upper-left quadrant — importance 4 or 5, evidence 1 or 2 — is the list that matters. (5) Rank the upper-left notes by importance, then by evidence; the top three to five get test cards (Chapter 9), in order. (6) Photograph and date the map; write the version note — which assumptions moved since the last map, and why. Deliverable: the dated map, the ranked list and three to five test cards. Pitfalls: scoring evidence from confidence rather than from the ladder; giving every assumption a 4 or a 5 on importance (if everything is critical, nothing is ranked — force a distribution); testing the upper-right assumptions because they are comfortable; and forgetting that assumptions move — the map is redrawn after every test, not once.

10.2.2 Dependencies among assumptions

Some assumptions cannot be tested until others are settled: whether owners renew at 80% cannot be known before a year of service, and whether a second bookkeeper can run the session cannot be tested before the session exists. The map therefore carries arrows — this assumption depends on that one — and the sequence of tests follows the arrows as well as the ranking. An important, unsupported assumption that cannot be tested this quarter gets a leading indicator instead: satisfaction after the first session and the "very disappointed" question stand in for renewal, and a test card is written for the indicator with an honest note that it is a proxy.

10.3 The Riskiest Assumption Test

The upper-left corner of the map holds the riskiest assumption — the one that would do the most damage and about which the least is known — and the discipline of this chapter is to test it first, with the cheapest test that can change the founder's mind, before anything else is built or bought.

KEY CONCEPT Test the assumption that kills you, not the one you can prove Founders test the assumptions they are confident about, because those tests pass, and the passing feels like progress. The assumption they avoid is the one whose failure would end the venture — that the segment will pay, that the channel exists, that the service can be delivered at the price — because a failure there feels like the end. It is the opposite: a killing assumption that fails in week six costs six weeks; the same assumption failing in month fourteen costs the venture. The map exists to make the avoidance visible. The rule is that no test is run on an assumption in the lower or right-hand quadrants while an assumption in the upper-left quadrant remains untested — and that the upper-left test is the cheapest one available, because the goal is to find out, not to be right.

10.3.1 Sequencing the tests

The sequence follows three rules. The riskiest first, as the key concept says. The cheapest test for it, from the ladder of Chapter 9: a page before a concierge, a concierge before a build. Dependencies respected: an assumption that cannot be tested until another is settled waits, with a leading indicator in the meantime. A quarter of discovery typically tests three to five upper-left assumptions, each with one or two turns of the loop, which is why the map must be short and the tests cheap. Ash Maurya described the whole discipline as moving from "plan A" — the first model — to "a plan that works", by attacking the riskiest parts of plan A in order (Maurya, 2012).

10.3.2 Discovery-driven planning

Rita McGrath and Ian MacMillan proposed, for ventures whose numbers are mostly assumptions, a planning method they called discovery-driven: start from the result the venture must achieve — the profit that would make it worthwhile — and work backwards through a reverse income statement to the revenue, the volume and the costs that result requires; list every assumption the statement rests on in an assumptions checklist; and tie each assumption to the milestone at which it will be tested, so that the plan is a sequence of checkpoints at which assumptions are converted into knowledge rather than a forecast to be met (McGrath & MacMillan, 1995). The method fits the discovery quarter exactly: the break-even and the three numbers of Chapter 7 are the reverse income statement; the assumptions table is the checklist; the test cards of Chapter 9 are the milestones; and the map of this chapter is what orders them. A founder who plans this way knows, at every milestone, which assumption is being bought with the next dollar.

10.3.3 What "tested" means

An assumption is tested when a test card with a threshold set in advance has been run to its end date and read. It is not tested by a conversation that went well, by a mentor's approval or by the founder's growing confidence, and it is not tested by a test whose threshold was adjusted afterwards. The validation log records, for each assumption, the cards run against it and their results; the map records its position; and the two together are what a lender, a co-founder or the founder's own future self will read.

10.4 Reading Results

A test ends with a number against a threshold, and the number is read in one of three ways. The reading is the moment when the discipline of the whole quarter is either kept or lost.

10.4.1 Validated, invalidated, inconclusive

Validated: the result met or exceeded the threshold, with a sample large enough and a source strong enough for the decision it supports; the assumption moves right on the map and is monitored. Invalidated: the result fell below the failure threshold with an adequate sample; the assumption is false as stated, and the model must change — which is the subject of Section 10.5. Inconclusive: the result fell between the thresholds, or the sample was too small, or the traffic did not fit, or the test broke; the assumption is unchanged, and the response is a stronger test — more traffic, a higher rung, a longer run — never a repeat of the same test in the hope of a better number. Table 10.1 presents the three readings with what each one means, the mistake most often made in reading it, and the next step. Read it as the decision table that closes every test card.

Table 10.1 — Reading a test result

10.4.2 Weighing the evidence

The strength of a result is the product of three things: the rung of the ladder (says, does, pays), the size of the sample (thirty is an anecdote, a hundred is a rate, five is enough for usability), and the source (friends, the founder's network, strangers from the segment). A validation from ten strangers' deposits outweighs one from fifty friends' survey answers, and the map's evidence score should reflect it. Daniel Kahneman's account of how people judge evidence — by how coherent a story it makes rather than by how much of it there is — is the warning: a single vivid interview makes a better story than a hundred sign-ups, and the founder must count rather than remember (Kahneman, 2011).

10.4.3 Confirmation and its remedies

Every founder reads results in favor of the model, and three remedies keep the reading honest. The threshold in advance, which this book has repeated in every chapter. The outsider — a mentor, an advisor, a co-founder who did not run the test — who reads the raw numbers before the founder's interpretation. And the learning card, which forces the founder to write what was observed before writing what it means, and which is filed whether the news is good or bad. A founder who finds that every test validates every assumption should re-read Chapter 3's warning about the interview that confirms everything: the tests are broken, not the model.

10.5 Persevere, Pivot or Stop

An invalidated assumption, or a set of results that together say the model cannot reach its thresholds, brings the founder to the decision this chapter exists for. Ries recommended a regular pivot-or-persevere meeting at which the evidence is reviewed and the decision is made deliberately rather than by drift (Ries, 2011); this section gives the decision its vocabulary and its criteria.

DEFINITION Pivot A pivot is a structured change to one part of the business model — the segment, the need addressed, the offer's scope, the channel, the revenue mechanism, the technology or the architecture — made in response to evidence, while keeping the parts of the model that the evidence has validated (Ries, 2011). A pivot is not a restart and not a random change: it names what changes and what stays, it is prompted by an invalidated assumption or a validated surprise, and it produces a new canvas and a new map whose riskiest assumption is then tested. The opposite of a pivot is not perseverance but drift — changing everything a little without deciding anything.

10.5.1 The ten kinds of pivot

Table 10.2 presents ten kinds of pivot, adapted from Ries's catalog. For each it states what changes, what stays, the signal that calls for it, and an example drawn from this book's fictional cases. Read the table when the evidence says the model must change but not how: most invalidated assumptions point to one or two kinds, and naming the kind keeps the change structured.

Table 10.2 — Ten kinds of pivot

10.5.2 Criteria for the decision

The decision is made on written criteria, set before the results arrive. Persevere when the upper-left assumptions have been validated on adequate evidence, the three numbers of Chapter 7 hold in the base case and the venture survives the pessimistic case, and the next test is a matter of scale rather than of existence. Pivot when an important assumption has been invalidated on adequate evidence, or a validated surprise — a segment that buys, a need the customers named, a feature they use — points to a better model that keeps most of what is validated; the pivot is named by kind, a new canvas is drawn, and its riskiest assumption is tested next. Stop when the riskiest assumption has been invalidated twice on strong evidence with no validated surprise to pivot toward; when no reachable segment with the pain has been found after the segments of the grid are exhausted; when the numbers cannot reach their thresholds under any plausible scenario; when the runway ends before the next test can run; or when the founder, honestly, applies the seventh kill criterion of Chapter 2 — they would not run this business for five years. A stop decided on criteria, with the log to show it, is not a failure; it is the cheapest outcome of a discovery quarter, and the founder who reaches it has three briefs in the log from Chapter 2 and a second one ready.

DO THE MATH A pivot decision by the numbers Illustrative figures. Daniel (Chapter 5) ran three paid pilots with construction contractors on his original offer — "your first hires done properly: contracts, onboarding, the rules" — at $1,200 each. The test card said: at least two of three will pay the regular price of $1,800 for a second hire within three months, and each will rate the paperwork as the main value. Result: one of three converted; all three rated "finding someone who showed up" as the value and the paperwork as "fine, I guess". The desirability assumption behind the offer was invalidated, and a validated surprise appeared. He drew the pivot in numbers. Persevere on the paperwork offer: contribution $900 per engagement, conversion 33%, so a first-year target of fifteen contractors requires forty-five pilots he cannot run — the weekly number of Chapter 5 fails. Customer-need pivot to "a crew of two hired and on site in two weeks, papers done": price hypothesis $2,400, contribution about $1,500 after his hours and the job postings, and, from the three pilots' words, an expected conversion above 60% — nine pilots for fifteen customers, feasible. Cost of the pivot: a new test card, a revised page, two more pilots — about twenty hours and $300. Cost of persevering: a quarter at the wrong conversion rate. He pivoted, recorded the decision with the three pilots' results in the registry, and kept everything the pilots had validated — the segment, the channel through the suppliers' counter, the relationship. The paperwork became a step inside the new offer, which is what a pivot is: a change to one block, with the validated blocks kept.

CASE IN POINT Three pivots of a scheduling application Fictional example; names and figures are invented. Créneau began as an online booking application for hair salons, built by a developer whose sister owned one. Pivot one, customer segment: after four months, the salons that used it were those whose owner was under thirty-five, and the larger, older-owned salons — the ones with revenue — would not switch from the phone; meanwhile, two physiotherapy clinics that had found the page were using it daily and asking for features. The developer redrew the canvas for clinics, kept the product and the channel, and grew to thirty clinics in a year. Pivot two, customer need: the clinics' interviews and the usage data showed that what they valued was not the booking but the reduction in no-shows — the reminders, the waiting list that filled a canceled slot — and that the booking itself competed with free tools; the offer was rewritten around "no-shows down by half", with the booking as a feature, and the price doubled without losing a clinic. Pivot three, value capture: a year later, the clinics' owners resisted the subscription in the slow summer months and three churned; a test of per-booking pricing on ten clinics showed a higher annual value and no churn, and the venture moved from a monthly fee to a small fee per completed appointment. Each pivot was prompted by an invalidated assumption or a validated surprise, each kept most of the model, each was recorded with its evidence — and none of them was visible in the founder's original plan, which had been a booking app for hair salons.

10.6 Risk Beyond Assumptions

Not every risk is an assumption about the model. Some risks are events that could damage the venture whether or not the model is right — a lawsuit, a supplier's failure, an illness, a rule change — and they are managed with a different instrument: the risk register.

10.6.1 The risk register

The register lists, for each risk the venture faces, its category, its likelihood and impact stated together in one column, the mitigation in place or planned, its owner and the trigger that would tell the founder it is materializing. The structure follows the general approach to risk management described in the international standard on the subject — identify, analyze, evaluate, treat, monitor — applied at the scale of a small venture (International Organization for Standardization, 2018). Table 10.3 gives the columns with example rows from the compost venture. Read it as a living document reviewed monthly, not as a form filled once for a lender.

Table 10.3 — The risk register (example rows from the fictional compost venture)

10.6.2 The founder's own exposure

The row that founders leave out is their own. A venture consumes the founder's savings, borrows against the founder's signature, occupies the founder's time and, when it fails, leaves the founder with the guarantees and the debts. The register includes the founder as a risk — their health, their capacity, their finances, their household — and the Québec and Canada Note gives the instruments that limit the exposure. Noam Wasserman's finding, that the choices founders make about control, equity and co-founders in the first months predict much of the later damage, belongs here as well: the founding team's own agreements are a risk row, and Chapter 12 gives the questions (Wasserman, 2012).

QUÉBEC & CANADA NOTE Personal risk: guarantees, incorporation and the founder's exposure (as of September 2026) A founder's exposure is decided by four things that Books A4 and A5 develop and that this note only names. The legal form. A sole proprietor is personally liable for the business's debts without limit; a shareholder of a corporation incorporated under Québec's Business Corporations Act, CQLR c S-31.1 or the federal Canada Business Corporations Act, R.S.C. 1985, c. C-44 is, in principle, liable only for the amount invested — subject to the exceptions that follow. Personal guarantees. Lenders, landlords and suppliers routinely require the founder's personal guarantee for a corporation's obligations, which removes the protection of the corporate form for that obligation; a guarantee should be read, limited in amount and duration where the lender will accept it, and recorded in the risk register — and a founder who is married or in a civil union should understand how family patrimony and the matrimonial regime affect the assets exposed, which a notary or a lawyer will explain. Directors' liabilities. A director of a corporation can be personally liable for certain unpaid amounts — unremitted source deductions and sales taxes, unpaid wages up to statutory limits — so that "incorporated" does not mean "insulated" from every debt. Insurance and income protection. General liability insurance for the venture's activities, professional liability where the offer is advice, and the founder's own disability and life coverage are the instruments that turn an event into a cost rather than a catastrophe; in Québec, self-employed workers are covered by the Québec Parental Insurance Plan and may opt into the federal Employment Insurance special benefits, under conditions that should be verified at the time. The practical rule of the discovery quarter is simple: sign nothing personal — a guarantee, a lease, a loan — before the riskiest assumptions of the map have survived their tests, and record every signature in the register with its amount and its term. Book A4 gives the legal forms and registrations; Book A5 the financing instruments and their guarantees; Book H1 the contracts; Collection G the taxes.

10.7 Deciding with Discipline

The decisions of this chapter — which assumption to test, how to read a result, whether to persevere, pivot or stop — are made by people who are tired, invested and hopeful. Three practices protect the decisions from the people who make them.

10.7.1 Decision records

Every decision of consequence gets a record: the question, the options considered, the evidence for each, the decision, the reasoning, the date, and the trigger that would reopen it. The record is short — half a page — and it is written before the decision is announced, because writing the reasoning is what tests it. A founder who keeps decision records can answer, a year later, the question every co-founder, lender and buyer asks — "why did you do it this way?" — with the evidence of the time rather than with a reconstruction. La Machine's registry of decisions, described in the closing section, is built for exactly this record, and its rule — no entry is closed without a written answer — is the discipline in software form.

10.7.2 The pre-mortem

Gary Klein proposed a simple exercise for the moment before a commitment: the team imagines that the plan has been carried out and has failed completely, and each member writes, independently, the reasons for the failure; the reasons are then read aloud and the plan is adjusted (Klein, 2007). The exercise works because it licenses the doubts that a hopeful group suppresses, and it takes fifteen minutes. In the discovery quarter, a pre-mortem is run before each commitment of money or time that the tests do not cover — a lease, a hire, a build, a launch — and its output goes into the risk register and the assumption map.

METHOD The pre-mortem (fifteen minutes) When to use it: before any commitment the tests do not cover — a lease, a hire, a build, a launch, a loan. Steps: (1) State the commitment in one sentence. (2) Say: "It is one year later. We did this, and it failed completely." (3) Each person, alone, writes for three minutes every reason the failure happened — no discussion, no filtering; a solo founder writes for five. (4) Read the reasons aloud, one per person in turn, until all are listed; group the duplicates. (5) For each reason, ask whether it is an assumption (it goes on the map), a risk (it goes on the register) or a design flaw (the commitment is changed). (6) Decide: proceed as is, proceed with the changes, or run one more test first. Deliverable: the list of reasons, sorted into map, register and changes, attached to the decision record. Pitfalls: running it as a discussion instead of as silent writing, which lets the loudest voice set the list; dismissing reasons as unlikely without recording them; and skipping it because "we've already decided" — which is exactly the moment it is for.

10.7.3 Sunk costs and the courage to stop

The strongest force against a good decision is the money and the time already spent. Hal Arkes and Catherine Blumer showed experimentally what every founder feels: having invested in something makes people more likely to continue investing in it, even when the future return is clearly worse than the alternatives (Arkes & Blumer, 1985). Kahneman placed the effect among the systematic errors of intuitive judgment, alongside overconfidence and the planning fallacy — the tendency to predict that the next stage will take less time and money than every previous stage did (Kahneman, 2011). The remedy is not willpower; it is the criteria written in advance, the outsider who reads the results, and the decision record that asks, explicitly, "what would we do if we were starting today, with what we now know, and nothing spent?"

WATCH OUT Sunk costs The fourteen months and the $62,000 of the Rideline case in Chapter 1 were not the worst of it; the worst was the six further months the founders spent "not wasting" the fourteen. Money and time already spent are gone whatever the founder decides next, and they are irrelevant to the decision — but they feel like a reason, and the feeling grows with the amount. The practical defenses are three: the stop rule written before the tests (Chapter 9); the pivot-or-persevere decision made at a fixed cadence on written criteria, not when the founder feels ready; and the question above, asked aloud by someone who has spent nothing. A founder who stops on criteria with a full log has lost a quarter and gained a method; a founder who cannot stop has lost the year that follows.

WHO DOES THIS Deciding in the three situations In a solo venture the founder makes every decision alone, which is why the criteria, the thresholds and the stop rule are written in advance and why one outsider — a mentor, an advisor, an experienced owner — reads the results before the founder interprets them; the decision record is written for the founder's own future reading. In a founding team, the pivot-or-persevere meeting is held at a fixed cadence with the evidence on the table, the decision rule is agreed before the meeting (consensus, or one owner's call after hearing the others), the pre-mortem is run as silent writing so that no member's optimism or pessimism sets the list, and the decision record names who decided and who disagreed. In an established business, the intrapreneur brings the map, the results and the recommendation to the gate, the owner or the general manager decides, and the decision is recorded with the same discipline as a capital investment — and the register includes the risk that the company's own habits smother the new model, which is real and rarely written. In all three, the registry of decisions is the venture's memory, and it is never edited to make a past decision look wiser.

Costly Mistakes

1. Testing the comfortable assumptions first. The page test on a feature the founder likes passes; the killing assumption — will they pay, can it be delivered at the cost — is tested in month fourteen. The consequence is the post-mortem's leading cause: a market that was never asked the fatal question.
2. Thresholds moved after the result. "Three percent is good for a first page"; "one of three is promising". The consequence is a quarter that validates everything and a model that rests on rationalizations, discovered by the bank account.
3. Rerunning an inconclusive test unchanged. The same page, the same traffic, a third time, hoping for a clearer number. The consequence is weeks spent on a question that only a stronger test could answer, and a founder who mistakes a small favorable sample for validation.
4. Drift instead of a pivot. A little change to the segment, a little to the price, a little to the channel, none decided, none recorded. The consequence is a model nobody can describe, a log nobody can read, and no way to know what was learned.
5. Signing personal guarantees before the riskiest assumptions are tested. A lease and a loan on the founder's signature for a model whose desirability is still an opinion. The consequence is a founder who cannot stop when the evidence says stop, because stopping now costs the house.
6. The sunk-cost continuation. Six more months to "not waste" the fourteen already spent. The consequence is the year after the failure, spent on the failure.

Key Takeaways

• Every model is a set of assumptions, extracted mechanically from the canvas, the assumptions table and the competitive response test, and sorted into four families of risk — desirability, feasibility, viability, adaptability — each with its own instruments.
• The assumption map ranks assumptions by importance and evidence; the upper-left quadrant — important and unsupported — holds the three to five that are tested first, and the map is redrawn after every test.
• Test the assumption that kills you, not the one you can prove, with the cheapest test that can change your mind, respecting dependencies and using leading indicators for what cannot be tested this quarter.
• Discovery-driven planning turns the plan into checkpoints: the required result, the reverse income statement, the assumptions checklist and the milestone at which each assumption is bought with the next dollar.
• A result is validated, invalidated or inconclusive, read against thresholds set in advance, weighed by rung, sample and source, and recorded on a learning card before it is interpreted; inconclusive calls for a stronger test, never the same one.
• Persevere, pivot or stop are decided on written criteria, and a pivot is a structured change to one part of the model, named by kind, that keeps what the evidence validated; a stop on criteria with a full log is the cheapest outcome of a discovery quarter.
• Risks that are not assumptions go in a risk register — category, likelihood, impact, mitigation, owner, trigger — reviewed monthly, with the founder's own exposure as a row and no personal signature before the riskiest assumptions have survived.
• Discipline is a set of practices, not a personality: decision records written before the decision, a pre-mortem before every commitment the tests do not cover, and criteria, an outsider and a fixed cadence to defeat sunk costs.

Your Action Plan

The plan below takes about six hours over the last weeks of the discovery quarter and produces the decision that Chapter 14's ninety-day plan ends with. Steps 1 and 2 build and rank the map; step 3 sequences the tests; step 4 reads the results; step 5 opens the register; step 6 makes and records the decision.

In La Machine

La Machine has no assumption map or risk register of its own, and it says so; what it has is the registry of decisions that this chapter's discipline requires, and the audit log, the registry of workflows and the runs, which show what changed, what is declared in production and what is waiting.

• Arbitrages (/arbitrages) is the registry of decisions: a question, the options, the recommendation, the decision and its date, kept forever — and, by its own rule, never closed without a written answer and never deleted because it aged or annoyed. The pivot-or-persevere decision, the pivot's kind and reasoning, the stop with its criterion, and every decision of consequence in the discovery quarter belong there; when a decision exceeds what the founder can settle alone — a co-founder's or a lender's call — the question is deposited and the work continues, which is the platform's rule of never stopping at a question.
• The audit log (/journal) records who changed what in the platform, and when — a task created, updated or completed, a run queued, skipped, approved or rejected, a project created or changed, a workflow's state changed — as the dated trail that the learning cards and the decision records refer to; a note added to a task writes no event there, so a test is recorded by closing its task with Mark as done.
• The registry (/registre), labeled Workflows, lists the venture's automations with their declared state — To build, Under construction, In production, Paused or Removed — and the same five states serve as a model for the assumptions of the map, which are, likewise, never "done" but always in a state that can be read.
• Tasks (/taches) carry each test card and each mitigation of the risk register as a task with its criterion and its date, so that the register is executed and not merely written; the date is set when the task is loaded, and the accountable person is named at the start of the task's objective, because the owner list of a task offers a single human actor.

What La Machine does not do: it does not score or map assumptions, weigh evidence, run pre-mortems or compute the value of a test — the map is drawn on a whiteboard tool or a spreadsheet attached to the project, and the register is a document reviewed monthly. Book A3 (Strategy and Competitive Positioning) develops scenario planning and strategic risk; Book A2 (Chapter 9) develops the risk register for a business plan; Book N2 shows the tasks and the audit log page by page, and Book N3 (Workflows, Automation and Human Approval) shows the registry of workflows.

Self-Check

1. Distinguish an assumption from a hypothesis, and name the four families of risk with the chapter that tests each.
2. What are the two axes of the assumption map, which quadrant is tested first, and how is the evidence score assigned?
3. Explain discovery-driven planning in three sentences and say what in this book plays the role of the reverse income statement and the assumptions checklist.
4. Define a pivot, name four of its ten kinds, and give the three written criteria for deciding to stop.
5. What is a pre-mortem, and what three practices defend a founder against sunk costs?

Answers. (1) An assumption is a belief the model depends on that evidence has not established; a hypothesis is its testable form — who, what, how many, threshold. Desirability is tested by Chapters 3 and 9, feasibility by Chapters 6 and 12, viability by Chapter 7, adaptability by Chapter 8 (Section 10.1). (2) Importance — would the model fail if the assumption were false? — and evidence — how much do we know, from which rung? The upper-left quadrant, important and unsupported, is tested first; evidence is scored 1 for the founder's belief up to 5 for what strangers paid (Section 10.2). (3) Start from the result the venture must achieve and work backwards through a reverse income statement to the revenue, volume and costs it requires; list every assumption the statement rests on; tie each to the milestone at which it will be tested, so that the plan is a sequence of checkpoints. In this book, the break-even and the three numbers of Chapter 7 are the reverse income statement and the assumptions table is the checklist (Section 10.3.2). (4) A structured change to one part of the model, made in response to evidence, keeping what the evidence validated; for example zoom-in, customer segment, customer need, value capture, channel. Stop when the riskiest assumption has been invalidated twice on strong evidence with no validated surprise; when no reachable segment with the pain exists after the grid is exhausted; when the numbers cannot reach their thresholds under any plausible scenario — or when the runway ends, or the founder would not run the business for five years (Section 10.5). (5) A fifteen-minute exercise in which the team imagines the plan has failed a year later and each member writes, alone, the reasons, which are then sorted into the map, the register and changes to the plan (Klein, 2007); the defenses against sunk costs are criteria and a stop rule written in advance, an outsider who reads the results, and a fixed cadence for the pivot-or-persevere decision (Section 10.7).

Chapter References

Business Corporations Act, CQLR c S-31.1. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/S-31.1

Canada Business Corporations Act, R.S.C. 1985, c. C-44. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-44/

Arkes, H. R., & Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and Human Decision Processes, 35(1), 124–140.

Bland, D. J., & Osterwalder, A. (2020). Testing business ideas: A field guide for rapid experimentation. Wiley.

International Organization for Standardization. (2018). ISO 31000:2018 — Risk management — Guidelines. ISO.

Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.

Klein, G. (2007). Performing a project premortem. Harvard Business Review, 85(9), 18–19.

Maurya, A. (2012). Running lean: Iterate from plan A to a plan that works (2nd ed.). O'Reilly Media.

McGrath, R. G., & MacMillan, I. C. (1995). Discovery-driven planning. Harvard Business Review, 73(4), 44–54.

Ries, E. (2011). The lean startup: How today's entrepreneurs use continuous innovation to create radically successful businesses. Crown Business.

Wasserman, N. (2012). The founder's dilemmas: Anticipating and avoiding the pitfalls that can sink a startup. Princeton University Press.

Chapter 11 — Typical Tasks, Cadence and Workflows

Chapters 1 to 10 gave the founder a method. This chapter gives it a calendar. The difference between a founder who reads about discovery and a founder who does it is not knowledge; it is a list of tasks with dates and criteria, a two-week rhythm that fits six hours a week, and a handful of automations that keep the logistics from eating the hours. This is that list, that rhythm and those automations, in the format that La Machine uses to run them.

What You'll Be Able to Do

• Describe the discovery rhythm — daily, weekly, the two-week sprint, monthly, and the gate at ninety days — and size it to your situation.
• Read and write a task in the La Machine format, with its ten fields and, above all, its verification criterion.
• Run the two-week discovery sprint day by day, within a six-hour weekly budget for a solo founder.
• Keep the validation log current with rules that survive a busy week.
• Hold the weekly review, the monthly model review and the go/no-go gate, each with its agenda and its output.
• Automate the seven workflows that carry the logistics of discovery — scheduling, intake, sign-ups, confirmations, digests, check-ins, withdrawals — and know what stays manual.
• Write verification criteria that make a discovery task delegable, automatable and countable.

Why It Matters

Discovery fails less often for lack of method than for lack of cadence. A founder with a day job and a family does not lack ideas about what to do; they lack a week in which the interviews were scheduled, the notes were filed, the count was taken and the decision was made — again, and again, for thirteen weeks. A task table with criteria turns the method into work that can be done on a Tuesday evening, delegated to a co-founder, or handed to a workflow; a cadence turns thirteen weeks into six sprints with a decision at the end of each. For a team, the format ends the question "who was supposed to do that?"; for an established business, it is the operating rhythm that a new offer's discovery must have before anyone approves a budget for it.

In This Chapter

• 11.1 The Discovery Rhythm
• 11.2 The Task Table in the La Machine Format
• 11.3 The Discovery Sprint
• 11.4 The Validation Log
• 11.5 The Model Review and the Go/No-Go Gate
• 11.6 Workflows Worth Automating First
• 11.7 Verification Criteria for Discovery Tasks

11.1 The Discovery Rhythm

Discovery has five cadences, nested like the rings of an operating rhythm, and each one holds a different kind of work.

11.1.1 Five cadences

Daily, ten to fifteen minutes: the founder checks what is due and what waits for a decision, logs every customer contact and every quote within the hour of having it, and approves or routes what the workflows have prepared. Weekly, three to four hours in the six-hour slot: the field work — interviews, tests, deliveries — and the count: contacts, conversations, sign-ups, deposits, hours. The sprint, every two weeks: a goal set on one assumption from the map, two rounds of field work, two syntheses, a review and a decision — the unit of progress that Section 11.3 describes day by day. Monthly, two hours: the model review — the canvas versioned, the numbers updated, the map redrawn, the risk register read, the runway recomputed — and the pivot-or-persevere question asked deliberately. The gate, at ninety days: the go/no-go decision of Chapter 10 on written criteria, recorded, and the next book opened.

11.1.2 Why cadence beats intensity

Founders want to compress discovery into a heroic fortnight, and the compression fails for three reasons. Interviews cannot be scheduled faster than customers accept them; a page test needs two weeks of traffic to reach its sample; a concierge test needs a delivery cycle. Synthesis needs distance: the founder who conducts six interviews on Monday and synthesizes them on Monday night sees what they hoped to see. And the day job, the family and the founder's own energy set a ceiling — six hours a week sustained for thirteen weeks produces seventy-eight hours of discovery; twenty hours a week for two weeks produces forty hours and an abandoned venture. The cadence is designed to be kept, and keeping it is the first verification criterion of the quarter.

KEY CONCEPT A discovery task without a criterion produces opinions "Talk to customers" is not a task; it is an intention that will be satisfied by any conversation, including one with a friend who liked the idea. "Conduct six thirty-minute interviews with owners above $300,000 by the 20th, notes filed within the hour, each with three verbatim quotes and one behaviour" is a task, because someone other than the founder — a co-founder, a mentor, the founder's own future self — can tell whether it was done. The criterion is what turns discovery from an activity into evidence; every row of Table 11.1 has one, and La Machine refuses to create a task without one. A founder who writes the criteria before the quarter has already done the most important synthesis of the method: they have said what "done" looks like.

11.2 The Task Table in the La Machine Format

La Machine describes every task with ten fields, so that a task can be created once, scheduled, assigned to a person or a workflow, executed and verified — the same format that Book B1 used for the marketing system and that a plan loaded into the platform carries, through its programming interface or a script, since no page creates a plan task. The ten fields are: Task (a verb and an object); Objective (the deliverable of the discovery quarter it serves — the problem, the segment, the proposition, the model, the numbers, the decision; or learning and compliance); Trigger (the event or date that starts it); Frequency (one-off, daily, weekly, fortnightly, monthly, quarterly, or event-driven); Automation level (Manual — a person does it; Semi-auto — the system prepares or prompts and a person decides or finishes; Auto — the system does it and a person reviews the log); Input (what the task needs to start); Output (what it produces); Verification criterion (how anyone can tell it was done correctly — observable, countable, time-bound); Role (who owns it — founder, co-founder, assistant, professional, or "system" for automated tasks); and Tools (the pages of La Machine or the tools of Chapter 13 used).

Table 11.1 lists the tasks of the discovery quarter as this book has built it: the twenty one-off setup tasks that produce the deliverables of Chapters 1 to 10 in dependency order, the twelve recurring tasks that keep the rhythm, and the seven workflows that automate the logistics. Every task refers to the chapter that explains it. Durations assume a solo founder on six hours a week; a team divides them, an established business staffs them. For legibility on a printed page, the ten fields are laid out in six columns, as in Book B1: the objective follows the task name in italics; the trigger, the frequency and the automation level share a column; the input and the output share a column with an arrow between them; the verification criterion has its own column; and the role and the tools share the last.

Table 11.1 — The discovery task table (format La Machine)

11.3 The Discovery Sprint

The sprint is the unit of progress: two weeks, one assumption from the map as its goal, two rounds of field work, two syntheses, and a review that ends in a decision and the next sprint's test cards, written in the test-card and learning-card format of the experimentation field guides (Bland & Osterwalder, 2020). The two-week box borrows from the design sprint, which compresses a problem, a prototype and a test with users into five days (Knapp et al., 2016), and from the build-measure-learn loop, in which every cycle ends in a decision to persevere or to pivot (Ries, 2011). Figure 11.1 lays it out over ten working days with the hours of a six-hour weekly budget beside each block.

Figure 11.1 — The two-week discovery sprint. Day 1 plans the sprint and recruits; days 2 to 4 run the interviews and the tests; day 5 is the first synthesis and the weekly count; days 6 to 8 run round two and the tests' second week; day 9 is the second synthesis and the map; day 10 is the review, the decision and the next cards — within six hours a week. Original figure.

11.3.1 Day by day

Day 1 — plan. The founder takes the top assumption from the map, writes or confirms its test card, and recruits: eight contacts, three sessions booked, the page or the concierge delivery of the fortnight prepared. One hour. Days 2 to 4 — field. Three interviews of thirty minutes with notes filed within the hour; the landing page live and its traffic sent; the concierge delivery of the week; every contact logged the same day; the Today page checked each morning. Three hours. Day 5 — synthesis one. The ninety-minute synthesis of Chapter 3 on the week's cards; the weekly count — contacts, conversations, sign-ups, deposits, hours — compared with the weekly number; three questions for round two and the script adjusted. Two hours. Days 6 to 8 — field, round two. Three more interviews, including two people who do not fit the brief; the page's second week; the pilot check-in; the next rung asked for in every conversation — a referral, a letter, a deposit. Three hours. Day 9 — synthesis two. The clusters merged across the two rounds; the map redrawn — what moved right, what was struck out; the learning cards for the tests that ended. Two hours. Day 10 — review and decide. Thirty minutes: results against thresholds, each test read as validated, invalidated or inconclusive; the decision recorded; next sprint's cards written. One hour.

11.3.2 Six sprints in ninety days

Thirteen weeks hold six sprints and a week of slack. The first two sprints are about the problem: the idea log and the briefs, then the interviews (Chapters 2 and 3). The third and fourth are about the model: the proposition, the sizing, the canvas, the numbers and the competitive map (Chapters 4 to 8). The fifth and sixth are about the riskiest assumptions: the page test, the concierge, the pilots, the map, the register and the gate (Chapters 9 and 10). Chapter 14 places the six on a calendar as the 30-60-90-day plan; a founder who keeps the sprint rhythm reaches the gate with a full log, whatever the gate decides. The order of the sprints — the customer and the problem before the product, the product before the model's numbers — is the order of customer development (Blank & Dorf, 2012).

METHOD Running the two-week discovery sprint When to use it: from the first week of discovery to the gate, six times. Steps: (1) On day 1, write the sprint goal as one sentence — the assumption to be tested and the threshold — and book the field slots in the calendar before anything else. (2) Run the field days with the rule of the hour: every note, contact and quote filed within sixty minutes of the event. (3) Hold synthesis one on day 5 with the cards, not from memory, and take the count before reading any result. (4) Adjust the script and the recruiting for round two from the three questions synthesis one produced. (5) Hold synthesis two on day 9, redraw the map, and write the learning cards before the review. (6) On day 10, review the results against the thresholds set on day 1, record the decision — including "no change" — and write next sprint's cards. Deliverable: per sprint, six filed notes, one or two test results, a dated map, one decision record and the next cards. Pitfalls: field days without booked slots ("I'll find time"); synthesis from memory; reading results before the count; a review that changes a threshold; and the most common — skipping day 10 because "nothing changed", which is exactly the decision to record.

DO THE MATH Hours per week by situation Illustrative figures. Table 11.2 sizes the recurring ring of Table 11.1 in hours per week for the three situations. For a solo founder: the daily check and logging, about half an hour a week; the field work of the sprint, three hours; the two syntheses and the count, spread, one and a half hours; the review, the map and the monthly items, spread, one hour — about six hours a week, which is the budget this book assumed in Chapter 1. Before the seven workflows, the same ring costs about nine hours, because scheduling, reminders, sign-up handling, confirmations, the digest and check-ins consume the founder's evenings; the workflows return about three hours a week — forty hours over the quarter, which is a full sprint's field work. A founding team of two runs two sprints in parallel and shares the reviews: about six hours each, twelve in total, and twice the evidence. An established business staffs the intrapreneur at fifteen hours a week — discovery is their job — and adds about five hours of the owner's, the accountant's and the sales representatives' time for the reviews and the gate: about twenty hours. In all three cases the hours are booked in the calendar before the quarter starts; a ring that lives in "when I have time" is not a ring.

Table 11.2 shows the weekly hours by cadence for the three situations. The rows are the cadences of Section 11.1; the columns are the situations; the last row is the total, and the note beneath states what the workflows return. Read it as a budget to book in the calendar rather than as an estimate of effort.

Table 11.2 — The recurring ring, in hours per week

Before the seven workflows, the solo ring costs about nine hours a week; the workflows return about three hours a week, or forty hours over the quarter.

11.4 The Validation Log

Chapter 3 gave the log its columns; this section gives it the rules that keep it alive through a busy quarter.

11.4.1 Four rules

The rule of the hour. Every observation — a quote, a behavior, a sign-up, a payment — is filed within sixty minutes of the event, because memory rewrites evidence into what the founder hoped to hear. One observation per row. A conversation produces several rows, one per thing said, done or paid; a row that summarizes a conversation is an opinion about it. Codes, not names. The shared version of the log identifies people by a code and a segment; the names live with the contacts, under the consent the rule of Chapter 3 requires. Nothing deleted. A struck-out hypothesis, a killed idea, a failed test stay in the log with their dates; the log is the record of what was learned, and what was learned includes what was wrong.

11.4.2 The weekly digest

Once a week — Friday, before the count — the log is read as a whole: new rows counted by hypothesis, the evidence board updated, and the three or four observations that changed something written into a short digest that the review reads on day 10. Workflow W5 assembles the raw counts; the founder writes the digest, because the judgment of what changed is the founder's. A founder who has kept the log and the digests for thirteen weeks has, at the gate, a document that a co-founder can join from, a lender can trust and the founder's own future self can read without embarrassment.

11.5 The Model Review and the Go/No-Go Gate

Three reviews punctuate the quarter, each with an agenda, a duration and an output. The weekly review is the sprint's day 10. The monthly model review and the gate are described here.

11.5.1 The monthly model review

Two hours on the first Monday of the month, with an outsider when possible, in six parts. The canvas: the current version read against the last, the version note written — what changed, on what evidence. The numbers: the assumptions table re-read, the three thresholds compared with the results, the runway recomputed. The map: redrawn from the month's learning cards; what moved right, what was struck out, what is next. The register: every trigger checked, new risks added, the founder's row current. The count: the month's weekly counts read as a trend against the weekly number. The question: pivot or persevere, asked deliberately and answered in the registry — including "persevere, no change", which is a decision and is recorded as one.

11.5.2 The gate

At day ninety, the go/no-go gate applies the criteria of Chapter 10 in writing and produces one of three outcomes. Go — persevere: the upper-left assumptions are validated on adequate evidence, the three numbers hold, the pessimistic case survives; the next book is A2 (the plan), A4 (the legal setup) or A5 (financing), depending on what the venture needs first, and its first action plan is dated. Pivot: an important assumption was invalidated or a surprise validated; the pivot is named by kind, the new canvas is drawn, and a second discovery quarter — shorter, because much is kept — starts on its riskiest assumption. Stop: a criterion of Section 10.5.2 was met; the log is closed with a summary, the second brief of Chapter 2 is opened, and the founder starts the next quarter with the method and without the loss. In an established business the gate is a meeting with the owner or the general manager, the intrapreneur presents the map and the results as numbers against thresholds, and the decision is recorded like a capital decision.

11.6 Workflows Worth Automating First

Discovery is manual work — conversations, judgments, first versions — surrounded by logistics that need not be. The seven workflows of Table 11.1 automate the logistics, and they are automated first because each one returns hours to the field work and removes a way of losing evidence.

11.6.1 The seven

W1 — interview scheduling and reminders. A scheduling link sends the accepted recruit to a calendar slot, a reminder the day before, and the consent statement before the session; the founder stops trading emails about Tuesdays. W2 — survey intake. A survey response is filed with its form, dated and coded, and creates a contact with its source and no consent, since only a lead magnet's consent box documents one, so that the count and the synthesis draw from one place. W3 — landing-page sign-up. A sign-up becomes a contact with the page as its source and the consent as given or not, receives an acknowledgment within minutes that identifies the sender and offers an unsubscribe, and creates a follow-up task — the acknowledgment being the first message of a relationship that Book B1 continues. W4 — pre-order confirmation. An order or a deposit produces a receipt that carries the disclosures of Chapter 3, records the deposit on its own ledger apart from operating cash, and states the delivery date and the cancellation conditions. W5 — the weekly digest. The counts of the week — contacts, sign-ups, responses, sessions — are compiled from the log, the forms, the pages and the calendar into the draft of the weekly count, which the founder confirms. W6 — pilot check-ins. Each cadence date of a pilot agreement sends a reminder to both parties and a feedback form whose responses go to the log. W7 — consent withdrawal. An unsubscribe or a withdrawal excludes the contact on the channel withdrawn, or on every channel the person names, from the next send and for every business of the account, and logs the reason, while a bounce reaches the exclusion list through a daily reading of the sending service's events — the workflow that Book B1 places at the center of the marketing system, installed here on the first day the venture holds a contact.

11.6.2 What stays manual

Three kinds of work are not automated in the discovery quarter, and a founder who automates them has automated the wrong thing. Conversations — interviews, observation, pilot meetings, the ask for a referral or a deposit — are the evidence; a form is not an interview. Judgments — the scorecard, the clusters, the reading of a result, the pivot-or-persevere decision — are the founder's, and an automation that "scores" them has removed the founder from the method. First versions — the brief, the profile, the statement, the canvas, the story — are written by the person who will defend them. What the workflows automate is the repetition of what a human decided: the reminder after the founder chose the cadence, the acknowledgment after the founder wrote it, the count after the founder defined it.

11.6.3 Approval

La Machine's assistant and agents apply one rule, and it is a good rule for a founder's own workflows: five actions — publishing, sending in bulk, spending, sending a proposal and deleting data — never happen without a human approval, while collecting, analyzing, qualifying and drafting proceed on their own. The workflows above respect it: W3 sends one acknowledgment to one person who asked for it; W4 confirms an order the customer placed; a bulk message to the waiting list, a published page or a proposal to a pilot waits for the founder's approval, which the Today page presents each morning. In the marketing module itself, as read on September 15, 2026, publishing a landing page and marking a sequence active are status changes rather than locks, so for those two actions the approval is the business's written rule, kept by the person who clicks (→ see Book N5).

11.7 Verification Criteria for Discovery Tasks

The criterion is the tenth field of the format and the one that makes the other nine work. This section says what a good one looks like.

11.7.1 Three properties

A verification criterion is observable — someone other than the person who did the task can see whether it was met, from a document, a count or a log; countable — it states a number, a share or a list, not an adjective; and time-bound — it says by when. "Interviews done" fails all three. "Six interviews conducted by the 20th, each with a note filed within the hour containing three verbatim quotes and one behaviour" passes them, and a co-founder, a mentor or a workflow can check it without asking the founder how it went.

11.7.2 Bad, better, best

Table 11.3 gives, for six discovery tasks, a criterion that fails, one that improves and one that passes. Read the rows as a training set: the pattern from left to right is always the same — an intention becomes a count, then a count with a condition and a date — and after six rows the founder can write the seventh alone.

Table 11.3 — Verification criteria for discovery tasks: bad, better, best

11.7.3 Criteria as the currency of delegation

A criterion is what lets a founder delegate without abdicating. A co-founder who receives "recruit twelve people" will recruit their friends; one who receives the best criterion of Table 11.3 will recruit strangers by segment and book the sessions, and the founder will know from the log whether it happened. A workflow that receives "send reminders" does nothing; one that receives "reminder at 24 hours, consent statement before the session, 100% of accepted recruits" can be built and checked. And an intrapreneur who reports "we talked to a lot of customers" to the gate is asked to come back; one who reports "twenty-four interviews across two segments, nine patterns, one invalidated feature, four deposits, here is the log" is asked what they need next. The criterion is the founder's own definition of done, written down where others can hold them to it.

WHO DOES THIS The rhythm in the three situations In a solo venture the founder owns every task in Table 11.1, the workflows are what make six hours enough, and the discipline is the booked slot — thirteen weeks of it, protected before the quarter starts — and the day 10 review held even when nothing changed. In a founding team, the setup tasks are split by the deliverables — one member owns the problem and the proposition, another the model and the numbers — the field work runs as two parallel sprints, the syntheses and the reviews are shared, and the criteria are what keep the split from becoming two ventures. In an established business, the intrapreneur owns the ring, the company's assistant or coordinator runs the workflows, the accountant joins the monthly review, and the owner holds the gate; the criteria are written into the mandate so that "discovery" is not a hobby the company tolerates but a project it can read. In all three, the tasks live in one system with their criteria, and the count is taken every week.

Costly Mistakes

1. Tasks without criteria. "Talk to customers", "test the page", "do the numbers" — satisfied by anything, verified by no one. The consequence is a quarter of activity and a gate with no evidence to decide on.
2. Intensity instead of cadence. Twenty hours in a heroic fortnight, then nothing. The consequence is forty hours and an abandoned venture, where six hours a week for thirteen weeks would have produced seventy-eight and a decision.
3. Synthesis from memory, results before the count. The interviews of Monday synthesized Monday night; the page's numbers read before the week's count. The consequence is evidence rewritten into hope, and a map that reflects the founder rather than the customers.
4. Automating the judgments and doing the logistics by hand. A tool that "scores" the interviews while the founder trades emails about Tuesdays. The consequence is a method with the founder removed from it and three hours a week lost to reminders.
5. Skipping day 10 because nothing changed. No decision recorded, no cards for the next sprint, the rhythm broken. The consequence is drift — the opposite of a pivot — and a log with no decisions in it.
6. A gate without criteria. "Let's see how we feel at ninety days." The consequence is the sunk-cost continuation of Chapter 10, decided by whoever is most attached, and a second quarter spent on the first quarter's question.

Key Takeaways

• Discovery has five cadences — daily, weekly, the two-week sprint, monthly, and the gate at ninety days — and cadence beats intensity because customers, tests and syntheses all take time, and because six hours a week for thirteen weeks is what a founder can actually keep.
• A task has ten fields in the La Machine format, and the verification criterion is the one that makes discovery produce evidence rather than opinions; the platform refuses a task without one.
• The task table holds twenty setup tasks in dependency order, twelve recurring tasks and seven workflows, each tied to the chapter that explains it and each with a criterion someone else can check.
• The sprint is the unit of progress: plan, field, synthesis, field, synthesis, review — six sprints fill the ninety days, and the founder reaches the gate with a full log whatever the gate decides.
• The log has four rules — the hour, one observation per row, codes not names, nothing deleted — and a weekly digest that the founder writes because the judgment of what changed is theirs.
• The monthly model review reads the canvas, the numbers, the map, the register and the count, and asks pivot-or-persevere deliberately; the gate applies the criteria of Chapter 10 in writing and opens the next book.
• Seven workflows automate the logistics first — scheduling, intake, sign-ups, confirmations, the digest, check-ins, withdrawals — and return about forty hours over the quarter; conversations, judgments and first versions stay manual; publishing, bulk sending, spending, proposals and deletions wait for a human.
• A good criterion is observable, countable and time-bound, and it is the currency of delegation: to a co-founder, to a workflow, to the gate.

Your Action Plan

The plan below takes about four hours in the first week of discovery and installs the rhythm that the other thirteen chapters assume. Step 1 books the time; steps 2 and 3 load the tasks and the criteria; step 4 installs the workflows; steps 5 and 6 fix the reviews and the gate in the calendar before the first sprint begins.

In La Machine

La Machine is built around the format of this chapter, and the discovery quarter runs on four of its pages.

• Tasks (/taches) hold every row of Table 11.1: the verification criterion is mandatory at creation — the product refuses a task without one — and a task has six statuses, of which active (recurring) is a healthy final state for the ring's tasks, not overdue work. No page creates a task: the rows are loaded through the programming interface or a script, and the assistant creates one at a time, born manual and without a due date; it can also list and count tasks, read one and push back a due date reversibly, although the postponement does not recompute the next due date, while dropping a task is deliberate and not undone. Projects (/projets) group the discovery quarter's tasks, one card per project with a gauge of completed and recurring tasks; no page creates a project or displays its documents, so the briefs, the log, the canvas versions, the sheets and the decision records are attached as document links to the tasks that use them.
• Today (/aujourdhui) is the daily cadence's screen, with four tiles and seven cards: the runs waiting for approval come first — nothing is more urgent, since the work of the assistant and the agents does not go out without it — then the overdue tasks, the postponed tasks whose date has arrived, the open work, the runs in motion, the recent failures and the blocked tasks. The work due today appears only as a count, and the list headed What you need to do follows one fixed owner rather than the person reading it, so the day's own list is read on Tasks sorted by due date. The guide's own caution applies to a founder's zeros — an overdue count of zero is good news, and a completed count of zero on a working morning deserves a look.
• Runs (/executions) show each occurrence of a plan task with its own status — queued, running, awaiting approval, done, failed, skipped — whether a person, the scheduler or a workflow wrote it, and awaiting approval is not done: the work exists, the human action that lets it go out has not been given; a workflow that runs without a plan task leaves no run there. The registry (/registre), labeled Workflows, lists the workflows with their state — To build, Under construction, In production, Paused or Removed — and In production is a declaration rather than a proof, since a workflow delivered is not a workflow running and the date of its last run is read in the runs.
• The approval rule — publishing, sending in bulk, spending, sending a proposal and deleting data require a human decision; collecting, analyzing, qualifying and drafting do not — governs the assistant and the agents, and the seven workflows as Section 11.6.3 described; in the marketing module, a published page or an active sequence is a status change that the business's written rule must guard (→ see Book N5).

One distinction matters: the platform holds two task systems that share a word and nothing else. The tasks of the discovery ring — recurring, with criteria, runs and approvals — live in Tasks and Runs, open to the administrator and agent roles; the customer work of the CRM — a meeting to hold, a follow-up to make, a pilot's check-in — lives in Productivity → Tasks (productivity/tasks), to which the older address crm/tasks redirects, as activities attached to a contact or a company, with no recurrence and no criterion. The two must not be confused: a row of Table 11.1 goes in the first, and the follow-up that W3 creates for one sign-up goes in the second, on the contact's record. What La Machine does not do: it does not write the criteria, take the interviews or hold the synthesis; the log and the map live in documents attached to the project. Book N2 (Piloting Tasks, Executions and Projects) covers these pages one by one, and Book N3 (Workflows, Automation and Human Approval) shows how the seven workflows are built and registered.

Self-Check

1. Name the five cadences of discovery and the work each one holds.
2. List the ten fields of a task in the La Machine format and say which one is mandatory at creation in the platform.
3. Describe the two-week sprint day by day, with the hours of a six-hour weekly budget.
4. Name the seven workflows automated first and the three kinds of work that stay manual, and state the platform's approval rule.
5. Rewrite "test the page" as a verification criterion with the three properties of Section 11.7.1.

Answers. (1) Daily — the check of what is due and the logging within the hour; weekly — the field work and the count; the sprint — a goal on one assumption, two rounds, two syntheses, a review and a decision; monthly — the model review; the gate at ninety days — the go/no-go decision (Section 11.1). (2) Task, objective, trigger, frequency, automation level, input, output, verification criterion, role, tools; the verification criterion is mandatory at creation (Section 11.2). (3) Day 1 plan and recruit, one hour; days 2–4 field, three hours; day 5 synthesis one and the count, two hours; days 6–8 field round two, three hours; day 9 synthesis two and the map, two hours; day 10 review and decision, one hour (Section 11.3). (4) Interview scheduling and reminders; survey intake; landing-page sign-up to contact and acknowledgment; pre-order confirmation; the weekly digest; pilot check-ins; consent withdrawal — and conversations, judgments and first versions stay manual; publishing, bulk sending, spending, proposals and deletions require a human approval, while collecting, analyzing, qualifying and drafting do not (Section 11.6). (5) For example: "Page live for fourteen days from the 3rd, at least 100 fitting visitors, sign-ups with a date counted, result read against the 8% threshold set on the card" — observable in the responses page, countable, time-bound (Section 11.7).

Chapter References

Blank, S., & Dorf, B. (2012). The startup owner's manual: The step-by-step guide for building a great company. K&S Ranch.

Bland, D. J., & Osterwalder, A. (2020). Testing business ideas: A field guide for rapid experimentation. Wiley.

Knapp, J., Zeratsky, J., & Kowitz, B. (2016). Sprint: How to solve big problems and test new ideas in just five days. Simon & Schuster.

Ries, E. (2011). The lean startup: How today's entrepreneurs use continuous innovation to create radically successful businesses. Crown Business.

Chapter 12 — Expertise, Roles and When to Get Help

A founder does not need to know everything. They need to know what they know, what they do not, which of the gaps must be filled by themselves, which by a co-founder, an advisor, a program or a professional — and what each of those costs, delivers and cannot be trusted with. This chapter is the honest inventory, and the guide to the help that surrounds a venture in Canada as of September 2026.

What You'll Be Able to Do

• Assess your own skills across six families on a grid you can share, and name the two that must stay in your hands whatever help you buy.
• Choose between going solo, founding a team and launching inside an established business, and settle the co-founder questions — roles, decisions, equity, vesting, commitment — before anything is built.
• Use advisors and mentors well: what each does, how to find them, and how to agree on scope, time and compensation without giving away the company.
• Read the incubators, accelerators and support organizations of the Québec and Canadian ecosystem, and pick the one that fits your stage.
• Hire professionals for the right deliverable at the right moment — accountant, lawyer or notary, designer, builder, researcher, financing advisor — with indicative fees and the questions that protect you.
• Brief and manage external help so that the work, the knowledge, the accounts and the intellectual property stay with the venture.
• Know what La Machine and artificial intelligence take over in the discovery quarter, and what remains a human's job.

Why It Matters

The post-mortems of Chapter 1 and Noam Wasserman's research on founders agree on a cause of failure that no method fixes: the people. A founder who cannot sell, a team that never settled who decides, an advisor with equity and no obligations, an agency that owns the accounts, a developer who kept the code — each is a venture killed by a role that was never defined. For the solo founder, this chapter is the map of the help that turns six hours a week into a whole venture. For a founding team, it is the conversation that prevents the split. For an established business, it is the mandate that lets an intrapreneur discover instead of being volunteered.

In This Chapter

• 12.1 Founder Skills
• 12.2 Solo Founder, Founding Team and the Established Business
• 12.3 Advisors and Mentors
• 12.4 Incubators, Accelerators and Support Organizations
• 12.5 Professionals
• 12.6 Briefing and Managing Help
• 12.7 What La Machine and AI Take Over

12.1 Founder Skills

The work of this book calls on six families of skill, and no founder holds all six. The grid of this section is the honest inventory; the rest of the chapter is what to do about the gaps.

12.1.1 Six families

Customer skills are the ability to find people who have the problem, to ask them about their lives without pitching, to hear answers one does not like, and to ask for a commitment — the whole of Chapters 2 to 5. Product and delivery skills are the ability to make or deliver the offer — the trade, the craft, the technology, the service — and to design the minimum version that tests it (Chapters 6 and 9). Financial skills are the arithmetic of Chapter 7 — costs, contribution, break-even, cash — and the habit of reading the numbers every week. Operational skills are the cadence of Chapter 11 — scheduling, logging, running the ring, keeping the promises the offer makes. Legal and compliance skills are the ability to recognize when a rule applies — consent, claims, pre-orders, taxes, the legal form — and to ask before rather than after. Leadership and people skills are the ability to recruit a co-founder, brief a professional, hold a difficult conversation with a partner, and make a decision on evidence when everyone in the room has an opinion.

12.1.2 The grid

Table 12.1 presents the six families with, for each, what a founder at level 1 and at level 5 can do, and who typically holds the skill in the three situations. The founder scores themselves from 1 to 5 in each family — and has the score checked by someone who has watched them work, because self-assessment in a founder's favorite family runs about a point high. Read the table as the input to every later section: a 1 or a 2 in a family is a gap to fill by a co-founder, an advisor, a program or a professional, and the last column says which.

Table 12.1 — Six skill families and who holds them in the three situations

12.1.3 The two skills that stay in the house

Whatever the grid says, two things are never bought. Customer knowledge — the interviews, the words, the patterns — belongs to the founder, because it is the venture's only real asset in the discovery quarter and because a founder who has not heard the customers cannot judge the advice of anyone who claims to speak for them. The decision — which problem, which segment, which model, persevere or pivot — belongs to the founder, because it is the venture, and because every advisor, program and professional has interests that are not the venture's. Everything else in the grid can be learned, shared with a co-founder, or bought by the hour.

12.2 Solo Founder, Founding Team and the Established Business

The first people decision is whether there will be other people. Wasserman's study of thousands of ventures found that founding teams both outperform solo founders on average and generate most of the conflicts that end ventures — and that the conflicts trace to decisions made, or avoided, in the first months about roles, control and equity (Wasserman, 2012). This section gives the questions; the answers are the founder's.

12.2.1 Solo

A solo founder has the whole grid to fill alone, and the whole decision to make alone. The method of this book was sized for them: six hours a week, the workflows, the outsider who reads the results. The gaps are filled by the hour — a professional, a builder — and by the mentor of Section 12.3, whose job is to be the disagreement the founder does not have at home. The risk of the solo founder is not capacity; it is the absence of anyone obliged to say "the interviews did not say that". The remedy is written thresholds, a decision registry that an outsider reads, and a mentor with a standing appointment.

12.2.2 The founding team

Two or three founders divide the grid and multiply the evidence, and they must settle six questions before the first sprint, in writing, because every one of them becomes a conflict when the venture succeeds or fails. Roles: who owns which deliverables of the discovery quarter, and which key activities of the canvas afterwards. Decisions: how a disagreement about the model is settled — consensus, one member's call after hearing the others, or a vote — agreed before the first disagreement. Equity: how ownership is split — not automatically equally, but by contribution, risk, capital and the value each brings, with the reasoning written down. Vesting: that each founder's shares are earned over time — the common practice is four years with a one-year cliff, so that a founder who leaves in month three does not keep a third of the company (Feld & Mendelson, 2019). Commitment: who is full-time, who keeps a job, and what changes when the venture can pay. Exit: what happens when a founder leaves, and how their shares are valued and bought — the substance of the shareholders' agreement that Book A4 and Book H1 develop. A team that has these six on one page, dated and signed, has removed the largest cause of failure that no test can catch.

WHO DOES THIS The founding team A solo founder is the team, and their first hire is usually a mentor, not an employee. A founding team settles the six questions of Section 12.2.2 in the first two weeks — the discovery quarter is the cheapest moment, because nothing has been built and nobody has yet sacrificed a year — and records them in a founders' agreement that a lawyer or a notary turns into a shareholders' agreement once the corporation exists; the team then splits the deliverables of Chapter 11 by the grid, not by preference, and holds the reviews together. An established business names an intrapreneur, gives them a written mandate — the segment, the budget, the hours, the gate criteria, the authority to say no to the sales force — and protects them from the company's urge to launch; the intrapreneur's "co-founders" are the owner who holds the gate and the outsider who reads the results. In all three, the person who owns customer knowledge and the person who makes the decision are named, and they are usually the same person.

12.2.3 Rich or king

Wasserman framed the founder's deepest choice as rich or king: founders who keep control — of decisions, of equity, of the title — tend to end up with a smaller share of a smaller venture, and founders who bring in co-founders, investors and experienced managers tend to own less of something larger (Wasserman, 2012). Neither is wrong; the error is not choosing. A founder who knows which they want makes the co-founder, advisor and financing decisions of this chapter and of Book A5 consistently, and one who does not makes them by mood, and resents the result.

12.2.4 The established business

An intrapreneur has advantages a founder lacks — customers, records, a brand, a sales force, an accountant — and one disadvantage that outweighs them: an organization built to run the existing model, which will absorb the new one into its habits unless someone prevents it. The mandate is the prevention: written, from the owner, with a budget line, protected hours, the gate criteria of Chapter 10, and explicit permission to refuse the shortcuts the company will offer — "just put it in the catalogue", "let the reps sell it", "use the existing price list". The intrapreneur's skills grid is read the same way as a founder's, and their gaps are filled from inside the company where the skill exists and from outside where the company's version would smother the model.

12.3 Advisors and Mentors

Advisors and mentors are the cheapest help a venture will get and the easiest to get wrong. This section separates the roles, says where to find them, and gives the agreement that keeps them useful.

12.3.1 Three roles

A mentor is an experienced person who meets the founder regularly, listens, challenges and shares pattern recognition — "I have seen this before, and here is what happened" — usually without compensation and without a formal role in the company; the mentor's value is the standing appointment and the disagreement. An advisor is a person with specific expertise or access — an industry, a technology, a market, a network — who contributes it on defined occasions, often under a short agreement and sometimes for a small equity grant or a fee. A board — of directors, or an informal advisory board — is a group that meets on a schedule, reviews the evidence and holds the founder to the decisions; a formal board of directors carries legal duties and belongs to a later stage (Book A4; Collection H), while an informal advisory board of two or three people is within reach of a discovery-stage venture and is often its best early governance.

12.3.2 Finding them

Mentors are found where experienced owners are: the association of the beachhead, the suppliers' counter, the accountants who refer, the programs of Section 12.4 — several of which exist to match founders with mentors — and the founder's own former employers and colleagues. Advisors are found by the gap: a founder with a 1 in the financial family looks for an owner who has run the numbers of a similar business, not for a famous name. The test of a candidate is the first meeting: a good mentor asks about the interviews before offering an opinion; a good advisor says what they do not know.

12.3.3 The agreement

An advisor relationship that involves equity, money or confidential information is written down on one page: the scope — what the advisor contributes, in which domain; the time — hours per month, meetings per quarter; the term — usually one year, renewable; the compensation — nothing, a fee, or an equity grant; the confidentiality and the ownership of anything the advisor helps produce; and the exit — how either party ends it. On equity, the practitioner norms that Brad Feld and Jason Mendelson describe apply: advisor grants are small — commonly a fraction of a percent to about one percent, depending on stage and contribution — and they vest over the term, so that an advisor who stops contributing stops earning (Feld & Mendelson, 2019). A founder should treat any advisor who asks for equity without vesting, or for a share that would be normal for a co-founder, as a warning sign, whatever the résumé.

WATCH OUT Advisors paid in equity without a vesting schedule The pattern is common and expensive: an impressive person agrees to "advise", the founder grants two percent of the company in gratitude, three meetings happen, and the advisor disappears — with two percent that every later investor will ask about and that no later contributor can recover. The consequence is dilution for nothing, a cap table that raises questions in every financing, and a lesson learned at the cost of the company's most scarce resource. The rule is simple: advisor equity is small, it vests over the term of a written agreement with defined contributions, and it stops vesting when the contributions stop. Book A5 (Financing a Business) treats the cap table and the dilution arithmetic; Book A4 the shareholders' agreement in which vesting lives.

12.4 Incubators, Accelerators and Support Organizations

Canada and Québec have a dense ecosystem of organizations that help ventures start, and a founder in discovery should know what each kind offers, what it asks in return, and which one fits the stage. Table 12.2 presents the kinds of organization with what they give, what they take or cost, the stage and the founder they fit, and the question to ask before joining. Read the table as a sorting device: a founder in the discovery quarter needs mentoring, evidence and a small budget, not office space and a demo day, and the kinds near the top of the table fit that need better than those near the bottom.

Table 12.2 — Kinds of support organization, what they give and what they ask (as of September 2026)

12.4.1 Choosing

Three questions decide which organization, if any, to join. What is the gap? A founder with a 1 in the customer family needs a mentor who has sold, not an accelerator; a founder with a validated model and a weekly number the channels cannot meet needs an accelerator's introductions or a program's financing, not more mentoring. What does it cost, truly? Equity given to an accelerator is a permanent cost; a program's time commitment is a cost measured against the six-hour slot; an incubator's space is a cost against a founder who could work from home. What do its alumni say? Every organization has founders who went through it; two conversations with them, asked with the script of Chapter 3 — what happened, what it cost, what they would do again — are worth more than the brochure. A founder in the discovery quarter usually joins nothing more than a mentoring network and a public advisory service, and keeps the accelerator question for the gate.

QUÉBEC & CANADA NOTE The support ecosystem for a founder (as of September 2026) The organizations below are named as of September 2026; programs change, and each should be verified on its own site before a founder counts on it. Mentoring: Réseau Mentorat matches Québec entrepreneurs with volunteer mentors for regular meetings (Réseau Mentorat, 2026); Futurpreneur Canada pairs its financing with up to two years of matched mentoring for founders aged 18 to 39 (Chapter 1; Futurpreneur Canada, 2026). Public advisory and financing: BDC offers advisory services alongside its financing and publishes free guides (Business Development Bank of Canada, 2026a); Investissement Québec finances and supports Québec businesses (Investissement Québec, 2026); Québec.ca's business pages and Canada.ca's business section are the official entry points for programs and registrations. Regional and local development: PME MTL in Montréal (PME MTL, 2026) and the local development organizations of other regions, and the SADC and CAE network in Québec's regions, offer advice and small loans; municipalities and regional county municipalities often run their own economic-development services. Training: the École des entrepreneurs du Québec offers courses for founders at every stage (École des entrepreneurs du Québec, 2026). Incubators and accelerators: university-linked incubators such as Centech (ÉTS) and District 3 (Concordia) in Montréal, the Creative Destruction Lab's Montréal site, Startup Montréal as the city's ecosystem hub, and, elsewhere in Canada, MaRS in Toronto and Communitech in Waterloo, among many others; accelerators such as Techstars operate cohorts in Canada, and some founders apply to programs outside the country. Professional orders: the Ordre des comptables professionnels agréés du Québec, the Barreau du Québec and the Chambre des notaires du Québec each maintain a public directory of their members, which is where a founder finds a professional whose title is verified. The Where to Get Help section of the master bibliography lists the addresses.

12.5 Professionals

Professionals are bought by the deliverable, at the moment the deliverable is needed, and the discovery quarter needs few of them. Table 12.3 presents the professionals a venture in discovery may engage, with their typical deliverables, indicative fee ranges in Canadian dollars as of 2026 — illustrative, varying by region, seniority and scope, and to be verified with two quotes before budgeting, with the federal job bank's published wages by occupation and region as a public benchmark for the hourly figures (Government of Canada, 2026) — the moment in the quarter when each is needed, and the controls that protect the venture. Read the table with Section 12.1.3 in mind: none of these professionals owns the customer knowledge or the decision, and the founder who hands either to a professional has bought the wrong thing.

Table 12.3 — Professionals for a venture in discovery (fee ranges illustrative, Canadian market, as of 2026 — verify before budgeting)

12.5.1 When not to hire

Three purchases are wrong in the discovery quarter, however tempting. A full product build before the concierge test — the $34,000 app of Chapter 9's case. A brand identity and a website before the proposition has passed the repeat test — the premature spending of Chapter 1's Watch Out. An agency or a consultant to "do the research" — the first twelve interviews are the founder's, and a report of opinions delivered by someone else is the bottom rung of the ladder at the top of the price list. The professionals of Table 12.3 are engaged for a defined deliverable, at the moment the method requires it, and the method requires very little before the gate.

DO THE MATH The cost of the quarter's help Illustrative figures. The discovery quarter of this book buys help sparingly, and the bill is small next to the year it prevents. A mentor, through a mentoring network: $0 to a few hundred dollars in fees, and two hours a month of the founder's time. An accountant: four hours across the quarter — the legal form, the numbers, the registration question — at $200, about $800. A lawyer or notary: two hours — the founders' agreement outline and the pre-order and consent language — at $350, about $700; the shareholders' agreement itself waits for the go decision. A designer: two days for the landing page and a clickable mock-up, about $1,400. A builder: none before the concierge; after it, a no-code single-feature MVP at 30 hours, about $2,700 — bought only if the gate says go. Tools: at the scale of a test, about $300 (Chapter 13). Total before the gate: about $3,200; with the MVP after it, about $5,900 — consistent with the $7,900 quarter of Chapter 1's Do the Math box, and less than a tenth of the untested year. A team pays the same; an established business pays it from a budget line the mandate created.

12.6 Briefing and Managing Help

Help is only as good as its brief, and a venture only keeps what its agreements say it keeps. This section gives both.

12.6.1 The brief

Every engagement — a designer's two days, a builder's thirty hours, a researcher's survey — starts from a one-page brief with seven parts. The objective: the assumption the work will test or the deliverable it produces, in the language of the method ("a landing page to test whether owners above $300,000 will leave an email at $180"). The context: the brief of Chapter 2, the profile of Chapter 4, the claims file — the professional works from the evidence, not from a phone call. The deliverable: exactly what is handed over, in what format, including source files and documentation. The criterion: how the founder will tell the work is done — the same three properties as Chapter 11. The budget and the deadline. The ownership: who owns the result and the accounts it lives in. And the confidentiality, where it matters. A professional who receives that page quotes accurately, delivers what was needed, and rarely surprises the founder.

METHOD Briefing external help in one page When to use it: before any engagement over a few hours. Steps: (1) Write the objective as the assumption or the deliverable, with its chapter. (2) Attach the context — the brief, the profile, the statement, the claims file — and say which parts the professional must respect verbatim (the claims) and which they may improve (the layout). (3) Describe the deliverable with its format, its source files and its documentation. (4) Write the criterion in the best form of Table 11.3. (5) State the budget as a cap and the deadline as a date, and ask for a fixed-fee quote where the deliverable is defined. (6) State the ownership and the accounts: everything created belongs to the venture, every account is opened in the venture's name with the founder as owner, and the professional's access can be revoked. (7) Add the confidentiality clause if the work touches a method, a customer list or a technology worth protecting, and skip it if it does not. Deliverable: the one-page brief, sent before the quote, attached to the task in the project. Pitfalls: briefing by phone; letting the professional rewrite the claims; accepting "I'll set it up under my account, it's easier"; paying the whole fee before the source files arrive; and the most common — a brief without a criterion, which produces work the founder cannot accept or refuse.

12.6.2 Ownership of what is made

Under the Copyright Act, R.S.C. 1985, c. C-42, the author of a work — a designer's page, a developer's code, a writer's text — is its first owner unless the work was made in the course of employment, in which case the employer owns it; a freelancer or an agency therefore owns what they make for the venture unless the contract assigns it in writing. The consequence, discovered too late by many founders, is a website the venture cannot modify, code it cannot take elsewhere, a logo it does not own. The rule of every brief is an assignment of copyright to the venture, signed before the work begins, with the source files delivered as part of the deliverable; Book H5 (Intellectual Property) gives the clause and its exceptions, and Chapter 8 gave the rest of the intellectual-property picture.

12.6.3 Accounts, credentials and data

The second thing a venture keeps is its accounts. Every tool, domain, page, payment account and data store used by a professional is created in the venture's name, with the founder as owner or administrator, and the professional granted an access that can be revoked — the rule that Book B1 stated for agencies and that applies with equal force to the builder who "sets up the workflows". Where the professional handles personal information — the researcher who moderates interviews, the builder who configures the contact base — the venture remains responsible for it under the privacy rules of Chapter 3, and the contract says so.

12.6.4 Managing the relationship

Three habits keep an engagement on track. One point of contact in the venture, who holds the brief. A weekly check against the criterion — not "how is it going?" but "what does the deliverable look like against the page we agreed?" — short, and written. And evidence over opinions in every disagreement: a designer who says the customers will prefer a different headline is asked what interviews say; a builder who says a feature is needed is asked which assumption it tests. The professional who accepts that discipline is worth keeping; the one who resists it is telling the founder something.

CASE IN POINT Two founders, one skills grid Fictional example; names and figures are invented. Norvale Compost's founders, two former kitchen managers, filled the grid of Table 12.1 in their second week, separately and then together. Both scored 4 or 5 on product and delivery — they had run kitchens and could run a route — and 4 on customer, because the restaurants were their former colleagues. Both scored 1 on financial, and both, when asked, had assumed the other would handle the numbers. Neither had thought about the legal form, the truck lease's guarantee or the permit the site would need. The grid produced four decisions, recorded in the registry. One founder took customers and routes, the other the site, the permits and the suppliers, and both took the interviews. They engaged an accountant for four hours a month from the third month — $800 for the quarter — who asked, in the first hour, what they had assumed for their own salaries, and rewrote the break-even from twelve customers to thirty-one. They asked a mentoring network for a mentor who had run a delivery business, and met him every three weeks; his first question was about route density, which became the unit-economics finding of Chapter 7. And they settled the six questions of Section 12.2.2 on one page — equal shares, four-year vesting with a one-year cliff, decisions by agreement with the customer-side founder's call on the model and the site-side founder's call on operations, both full-time after the gate if it said go — which a notary turned into a shareholders' agreement for $3,500 when they incorporated. The grid had cost them an hour. The accountant's first hour had changed the break-even by a factor of two and a half.

12.7 What La Machine and AI Take Over

The last section of the inventory is the machine's share. In the discovery quarter, La Machine and the AI assistants of Chapter 13 take over a specific set of tasks, and the boundary of that set is as important as its contents.

12.7.1 What the machine takes over

Drafting first versions from the founder's material. An interview script from the brief; a landing page's copy from the statement and the claims file; a survey's questions from the interviews' words; the seven sentences of the story from the canvas. The drafts are the founder's to correct, and the material they draw on is the founder's evidence, not a blank prompt. Summarizing and structuring the founder's notes. Clustering the quotes of a round, extracting the behaviors from the notes, drafting the learning card — with the caution below about personal information. Computing. The chains of Chapter 5, the unit economics and the scenarios of Chapter 7, the value of a test of Chapter 9; the calculators of La Machine and a spreadsheet do the arithmetic, and the founder supplies and defends the assumptions. Running the logistics. The seven workflows of Chapter 11 — scheduling, intake, sign-ups, confirmations, the digest, check-ins, withdrawals — which the platform runs under its approval rule. Monitoring. The Today page, the runs, the registry: what is due, what waits, what failed.

12.7.2 What remains human

The conversations: no assistant interviews a customer, watches them work, or asks them for a deposit. The judgments: the scorecard, the clusters' meaning, the reading of a result against its threshold, the persevere-or-pivot decision — the assistant can lay out the options, and the founder decides and records. The first versions that will be defended: the brief, the profile, the statement, the canvas; the assistant drafts, the founder writes. The relationships: the co-founder conversation, the mentor's challenge, the professional's brief, the pilot customer's trust. And the responsibility: for the claims on the page, for the personal information collected, for the decision at the gate — the platform records, and the founder answers.

12.7.3 A caution on personal information and AI tools

The notes of twelve interviews contain personal information — names, circumstances, sometimes financial details — collected under the consent statement of Chapter 3. Pasting them into a third-party AI assistant to "summarize" sends that information to a service the interviewees were not told about, which engages the purpose-limitation and consent rules of Québec's Act respecting the protection of personal information in the private sector, CQLR c P-39.1 and, where the service stores data outside Québec, its rules on communication outside the province. The practical rule is to summarize from coded, de-identified notes — the log's shared version, with names replaced by codes — to check the AI tool's terms on what it retains and where, and to name the tool in the consent statement when it will be used; Book H2 (Privacy and Personal Information) and Book K4 (Artificial Intelligence for Business) give the full treatment, and Book N11 describes how La Machine's own assistant works on the venture's data within the platform rather than on a blank prompt.

Costly Mistakes

1. A skills grid never filled, or filled a point high. Two founders who each assumed the other would handle the numbers; a solo founder who believed a 3 in the customer family was a 4. The consequence is a gap discovered by its cost — a break-even off by a factor of two and a half, or twelve interviews that were pitches.
2. The co-founder questions avoided. Roles by habit, equity "we'll sort out later", no vesting, no decision rule. The consequence, in Wasserman's data and in every accountant's experience, is a venture that succeeds and splits, or fails and litigates.
3. Advisor equity without vesting. Two percent for three meetings and a name on a slide. The consequence is dilution for nothing and a cap table that every later investor questions.
4. The wrong organization for the stage. An accelerator's equity given for introductions a founder in discovery cannot use; an incubator's rent paid for a desk the founder does not need. The consequence is a permanent cost for a temporary need.
5. The research, the build or the brand bought before the method requires them. An agency's report of opinions, a $34,000 app, a brand identity for a proposition that failed the repeat test. The consequence is the untested year of Chapter 1, paid for with the professionals' invoices on top.
6. Work, accounts and data owned by the supplier. A website the venture cannot edit, code it cannot take, a page under the designer's account, interview notes pasted into a tool nobody was told about. The consequence is a venture that must start again, or pay to stay — and, for the notes, a privacy breach on top.

Key Takeaways

• Six families of skill — customer, product and delivery, financial, operational, legal and compliance, leadership and people — are scored honestly on a grid, checked by someone who has watched the founder work; a 1 or a 2 is a gap to fill by a co-founder, an advisor, a program or a professional.
• Two things are never bought: customer knowledge and the decision. Everything else can be learned, shared or hired by the hour.
• The founding team settles six questions in writing before the first sprint — roles, decisions, equity, vesting, commitment, exit — because every one of them becomes a conflict when the venture succeeds or fails; the intrapreneur needs a written mandate for the same reason.
• Mentors challenge on a standing appointment; advisors contribute expertise or access under a one-page agreement with scope, time, term and small, vesting equity or none; a board comes later, and an informal advisory board of two or three people can come now.
• Support organizations are sorted by the gap they fill and the cost they ask — mentoring networks and public advisory services fit the discovery quarter; accelerators and their equity wait for a validated model and the gate.
• Professionals are bought by the deliverable at the moment the method requires it, with fixed-fee quotes for defined work, indicative fees verified by two quotes, and never for the first twelve interviews, a full build or a brand before the repeat test.
• A one-page brief with objective, context, deliverable, criterion, budget, ownership and confidentiality is what keeps an engagement on track; copyright is assigned in writing, accounts are in the venture's name, and personal information stays under the venture's responsibility.
• The machine drafts, summarizes, computes, runs the logistics and monitors; the human converses, judges, writes the first versions that will be defended, holds the relationships and answers for the result — and de-identified notes are the only ones that go into an AI tool.

Your Action Plan

The plan below takes about five hours in the first three weeks of the quarter and puts the people around the venture in place before the sprints need them. Step 1 fills the grid; step 2 settles the team; steps 3 and 4 find the mentor and the organization; step 5 engages the two professionals the quarter needs; step 6 writes the brief template that every later engagement will use.

In La Machine

La Machine holds the people around the venture as records, runs the assistant and the agents within the approval rule, and leaves the human decisions to the founder.

• Accounts and roles decide who works inside the platform: a co-founder, an assistant or a professional given access receives an account whose product role — administrator, agent or another — decides which pages open, and that access should be granted for the work, recorded as a task with its end date and removed when the work closes, which is how the accounts-in-the-venture's-name rule of Section 12.6.3 is applied inside the platform. The Agents page (crm/agents), which has no menu entry, is not where people are added: an agent there is a line of business rather than a person, its Add Agent button is disabled, and a representative is a contact.
• The assistant (/agents-ia) and orchestration (orchestration) are where the platform's assistant and its agents run the work that Section 12.7.1 lists — drafts, summaries and digests, with the agents called by the workflows — under the rule that publishing, sending in bulk, spending, sending a proposal and deleting data wait for a human approval; the assistant works on the venture's own data within the platform, from a closed catalog of capabilities, which is the design that Book N11 describes and that the caution of Section 12.7.3 requires.
• Contacts and companies (business/marketing/contacts, business/marketing/companies) hold the mentor, the advisors, the professionals and the organizations as records with their agreements' terms in the notes, so that the advisory agreement's term and the professional's deliverable are dated where the founder will see them.
• Organizations (crm/organizations), which has no menu entry and opens from the organization switcher, lists the workspaces a user belongs to; as of September 2026 it reads a model built for the platform's future multi-tenant accounts, which does not hold the account's own organization, so a founder should not read that screen as a description of the venture's structure.

What La Machine does not do: it has no directory of mentors, advisors, incubators or professionals — the Québec and Canada Note and the master bibliography's Where to Get Help section name them, and the professional orders' directories verify titles — and it does not draft or hold the founders' agreement, the advisory agreement or the shareholders' agreement, which are the founder's documents with a lawyer or a notary. Book N11 (The AI Assistant, Agents and Orchestration) describes the assistant and the agents page by page; Book K4 (Artificial Intelligence for Business) gives the rules for using AI tools with business and personal data; Book A4 and Book H1 give the agreements.

Self-Check

1. Name the six skill families and the two things that are never bought, and explain why.
2. What are the six questions a founding team settles before the first sprint, and what does vesting with a cliff protect against?
3. Distinguish a mentor, an advisor and a board, and give the practitioner norm for advisor equity.
4. Which kinds of support organization fit the discovery quarter, which one waits for the gate, and what three questions decide?
5. Name the seven parts of a one-page brief, and state what the Copyright Act implies for work made by a freelancer.

Answers. (1) Customer; product and delivery; financial; operational; legal and compliance; leadership and people. Customer knowledge and the decision are never bought, because the first is the venture's only real asset in discovery and the second is the venture itself — and every helper has interests that are not the venture's (Section 12.1). (2) Roles, decisions, equity, vesting, commitment and exit; vesting over time with a cliff protects against a founder who leaves early keeping a large share of the company (Section 12.2.2). (3) A mentor meets regularly, listens and challenges without a formal role; an advisor contributes specific expertise or access on defined occasions under an agreement; a board reviews evidence on a schedule and holds the founder to decisions, with a formal board carrying legal duties later; advisor equity is small — a fraction of a percent to about one percent — and vests over the agreement's term (Section 12.3). (4) Mentoring networks and public advisory services fit the discovery quarter; accelerators, with their equity, wait for a validated model and the gate; the questions are what gap it fills, what it truly costs, and what its alumni say (Section 12.4). (5) Objective, context, deliverable, criterion, budget and deadline, ownership, confidentiality; under the Copyright Act, the author of a work is its first owner unless made in the course of employment, so a freelancer's page, code or logo belongs to the freelancer unless the contract assigns it to the venture in writing (Section 12.6).

Chapter References

Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25). LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1

Copyright Act, R.S.C. 1985, c. C-42, s. 13. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/C-42/

Business Development Bank of Canada. (2026a). Advisory services. Retrieved September 7, 2026, from https://www.bdc.ca/en/consulting

École des entrepreneurs du Québec. (2026). Formations. Retrieved September 7, 2026, from https://eequebec.com/

Feld, B., & Mendelson, J. (2019). Venture deals: Be smarter than your lawyer and venture capitalist (4th ed.). Wiley.

Futurpreneur Canada. (2026). Core startup program. Retrieved September 7, 2026, from https://futurpreneur.ca/en/offering/core-startup/

Government of Canada. (2026). Job Bank: Explore careers — wages. Retrieved September 7, 2026, from https://www.jobbank.gc.ca/explorecareers

Investissement Québec. (2026). Financing and support. Retrieved September 7, 2026, from https://www.investquebec.com/

PME MTL. (2026). Services aux entreprises. Retrieved September 7, 2026, from https://pmemtl.com/

Réseau Mentorat. (2026). Le mentorat pour entrepreneurs. Retrieved September 7, 2026, from https://reseaumentorat.com/

Wasserman, N. (2012). The founder's dilemmas: Anticipating and avoiding the pitfalls that can sink a startup. Princeton University Press.

Chapter 13 — Tools: Functions and Popular Tools

The discovery quarter needs fewer tools than founders buy, and most of them are free at the scale of a test. This chapter lists what each function must do — which does not change — and then the popular tools that do it, with their pricing model as of September 2026 — which changes every quarter. Read the functions as the specification and the tools as a starting list of names; then verify, on the vendor's page, what a tool costs and includes today.

What You'll Be Able to Do

• Describe the discovery stack by function — ten functions, from canvases to workflow automation — and say what each must do before naming a tool.
• Apply four selection rules: free and already-paid first, connect everything, one source of truth, know where the data lives.
• Choose tools function by function from dated tables, including La Machine where it fits, without paying for what a test does not need.
• Draw the integration map of the stack around one hub, and find the breaks where evidence is lost.
• Compute the total cost of the discovery stack in dollars and in hours, and keep a stack inventory that a later book will inherit.
• Place La Machine in the stack — what it covers, what it does not, how it connects.

Why It Matters

Tools are where founders spend the discovery quarter's money and lose its evidence. The money goes on subscriptions bought "to be ready" for a volume that never arrives; the evidence goes into a notes application on one laptop, a form whose responses nobody exports and a second contact list that the workflows never see. A stack chosen by function, connected around one hub and inventoried from the first day costs a few hundred dollars for the quarter and holds every interview, sign-up and deposit where a decision can read them. For a team, one source of truth ends the question "where is the latest version?"; for an established business, it keeps the new offer's evidence out of the company's systems until the gate says it belongs there.

In This Chapter

• 13.1 The Discovery Stack by Function
• 13.2 Selection Rules
• 13.3 Function-by-Function Tables
• 13.4 The Discovery Integration Map
• 13.5 The Total Cost of the Discovery Stack
• 13.6 La Machine in the Stack

13.1 The Discovery Stack by Function

Ten functions cover the work of this book. Each is described by what it must do — the "must do" list that a tool is measured against — and the descriptions are stable across years, which the tools are not.

13.1.1 Ten functions

Canvases and whiteboards hold the boards of Chapters 4, 6 and 10 — the profile and value map, the nine-block canvas, the assumption map — as movable notes that can be marked, versioned and photographed. Interview notes, recording and transcription capture the conversations of Chapter 3 within the hour, with consent, and turn them into quotes on cards. Surveys and forms collect responses and sign-ups with the unbundled consent of Chapters 3 and 9. Market, trend and competitor data supply the counts of Chapter 5 and the alternatives of Chapter 8. Landing pages and smoke tests publish the test pages of Chapter 9 and count their visitors and responses. No-code prototypes and MVPs build the mock-ups, the Wizard of Oz fronts and the single-feature products of Chapter 9 without a developer. Financial modeling and calculators compute the chains, the unit economics, the scenarios and the runway of Chapters 5 and 7. The validation log, contacts and CRM are the hub of Chapter 11 — every observation, every person, every consent, every source, in one place. Scheduling and payments book the interviews and the pilots and take the deposits of Chapters 3 and 9, within the consumer rules. AI assistants and workflow automation draft, summarize, compute and run the seven workflows of Chapter 11, under the approval rule and the privacy caution of Chapter 12.

13.1.2 Functions first, tools second

The order matters for a reason beyond tidiness. A founder who starts from a tool inherits the tool's idea of the work — a survey tool makes every question a survey question, a page builder makes every test a page — and a founder who starts from the function chooses the cheapest tool that does what the method requires and nothing more. The tables of Section 13.3 therefore open with the function's "must do" list, and every tool in them is measured against that list, La Machine included.

13.2 Selection Rules

Four rules, and four questions before any purchase, keep the stack small and the evidence whole.

13.2.1 Four rules

Free and already-paid first. Most functions of discovery are covered by a free tier or by a tool the founder already pays for — the office suite's forms and spreadsheets, the calendar's booking pages, the notes application, the platform subscription; a paid tool is adopted when a task of Chapter 11 requires what the free one cannot do, and not before. Connect everything. A tool that does not send its data to the hub — a form whose responses stay in the form, a page whose sign-ups stay on the page — creates a break where evidence is lost; Section 13.4 draws the map. One source of truth. The log holds every observation and the contact base holds every person, consent and source; every other tool feeds them or reads them, and a second list anywhere is a bug. Know where the data lives. Personal information collected in discovery lives in tools whose terms and locations the founder has read; the Québec and Canada Note of Section 13.5 gives the rule.

13.2.2 Four questions before buying

Before any subscription, four questions, and a written answer to each. Which task of Table 11.1 requires it? If none, the tool waits. What does it replace? A tool that duplicates a function already covered adds a list, not a capability. Who owns the account? The venture, with the founder as owner and every collaborator's access revocable — the rule of Chapter 12. What happens to the data when we leave? Export at any time, in a usable format, and deletion on request; a tool that answers "you can't" to either has answered the question.

13.3 Function-by-Function Tables

The ten tables that follow list, for each function, its "must do" and then five to eight popular tools with the function they cover, what they are best for, their pricing model as of September 2026 and a note. No amount is printed, because amounts change several times a year and this library cites prices only from a dated vendor page; the model — free, freemium, per user, per volume, per transaction, included in a subscription — changes less often and is what a founder needs to compare. La Machine appears in the function it covers, described and not recommended; the alternatives are described in the same neutral terms. Every table is a starting list, not a ranking.

13.3.1 Canvases and whiteboards

A canvas tool must keep movable notes, one item each, in colors; mark evidence and hypothesis differently; let a board be duplicated for a version and photographed or exported; allow sharing with a co-founder or an outsider; and export the board's contents. Table 13.1 lists tools for this function; the last row is the one most founders forget.

Table 13.1 — Canvases and whiteboards (as of September 2026)

13.3.2 Interview notes, recording and transcription

A note-taking or transcription tool must let notes be filed within the hour; record with consent stated at the start; produce a transcription that can be corrected; extract quotes as cards; store the data where the founder controls it and can delete it; and show codes instead of names in the shared version. Table 13.2 lists the tools; the privacy caution of Chapter 12 applies to every row that transcribes.

Table 13.2 — Interview notes, recording and transcription (as of September 2026)

13.3.3 Surveys and forms

A survey tool must run short surveys with closed and open questions; carry a consent checkbox that is separate, unchecked, and names who will write and about what; export responses with dates; offer a sign-up form for the waiting list; serve French where required; and connect to the contact base. Table 13.3 lists the tools; the consent design of Chapters 3 and 9 is the founder's responsibility on every row.

Table 13.3 — Surveys and forms (as of September 2026)

13.3.4 Market, trend and competitor data

A market-data source must give counts of businesses and people by industry, size and region; search interest by term, region and time; competitors' public materials, reviews and filings; a countable list of names in the segment; and dated observations. Table 13.4 lists the sources; the first rows are free, official and the ones this book cites, and the national statistical agency's portal is the entry point to its counts, profiles and releases (Statistics Canada, 2026b).

Table 13.4 — Market, trend and competitor data (as of September 2026)

13.3.5 Landing pages and smoke tests

A landing-page tool must publish one page per offer or price in minutes; carry a form with unbundled consent; count visitors and responses; hold a draft state until a person approves publication; serve French where required; and respect the truthful-claims discipline of Chapter 4 on every page. Table 13.5 lists the tools; the first row refuses drafts at its public address, and the approval of each publication is the business's written rule.

Table 13.5 — Landing pages and smoke tests (as of September 2026)

13.3.6 No-code prototypes and MVPs

A prototyping tool must produce a clickable mock-up for five users; a Wizard of Oz front — a form, a page, a message — with the work done behind it; a single-feature product assembled without code; data in a table the founder owns; and an export when the venture moves on. Table 13.6 lists the tools; the first row says what the platform does and does not do.

Table 13.6 — No-code prototypes and MVPs (as of September 2026)

13.3.7 Financial modeling and calculators

A financial-calculator tool must hold the chains of Chapter 5; the unit economics, break-even, cash and scenarios of Chapter 7; an assumptions table with ranges and sources; formulas the founder can read; and a version a lender can open. Table 13.7 lists the tools; most founders need the first two rows and nothing else in the quarter, and a free plan template gives the structure when the plan of Book A2 begins (Business Development Bank of Canada, 2026c).

Table 13.7 — Financial modeling and calculators (as of September 2026)

13.3.8 The validation log, contacts and CRM

A log and contact tool must keep one observation per row, with date, source, method, hypothesis, rung, weight and implication; one record per person, with segment, source and consent state; enforce the do-not-reach list; export at any time; give access to a co-founder or an outsider; and show codes instead of names in the shared view. Table 13.8 lists the tools; the hub of Section 13.4 is chosen here.

Table 13.8 — The validation log, contacts and CRM (as of September 2026)

13.3.9 Scheduling and payments

A scheduling and payments tool must offer a booking link for interviews and pilot meetings, with reminders; write records to the contact base; take deposits by a means that preserves the consumer's chargeback right; keep a deposits ledger apart from operating cash; and issue receipts that carry the disclosures of Chapter 3. Table 13.9 lists the tools; the consumer rules decide the payment rows.

Table 13.9 — Scheduling and payments (as of September 2026)

13.3.10 AI assistants and workflow automation

An assistant or automation tool must draft from the founder's own material; summarize de-identified notes; check calculations; run the seven workflows of Chapter 11 with approval points; log every run; and handle data under the privacy rules of Chapters 3 and 12. Table 13.10 lists the tools; the caution of Chapter 12 applies to every assistant row, and the approval rule to every automation row.

Table 13.10 — AI assistants and workflow automation (as of September 2026)

TOOLS OF THE TRADE The five tools of discovery on day one (as of September 2026) Before any of the tables above, five things, most of them free or already paid: a place for the validation log — a spreadsheet with the columns of Table 3.2, or a database, attached to the venture's project; a form with an unbundled consent checkbox — the platform's forms, or a free form tool configured with the wording of Chapter 3; a scheduling link — the calendar's own booking page, or a freemium scheduler connected to the contacts; a spreadsheet for the numbers — the sizing sheet and the assumptions table, shared with the accountant; and a contact base that records consent and source for every person — the platform's contacts, or a free-tier CRM configured for it. When the page test of Chapter 9 comes, a sixth is added — a landing page tool, or the platform's pages — and when the concierge becomes an MVP, a seventh. Everything else waits for a task of Table 11.1 that requires it. Five tools, all but one free or already paid, cover the discovery quarter of a solo founder.

WHAT CHANGED Prices change: how to read this chapter in 2027 The pricing models in the tables — free, freemium, per user, per volume, per transaction, included in a subscription — change less often than the amounts, which is why no amount is printed here and why the library's rule is to cite a price only from a dated vendor page. Between 2023 and 2026, several vendors in these tables moved features from free to paid tiers, added AI features to higher tiers, changed the limits of their free plans, raised the cost of automation beyond a threshold of tasks or executions, and were acquired or renamed; the assistants' tiers changed more often than any other row. Read each table as a map of the function and a starting list of names; then verify, on the vendor's pricing page and with the date recorded in the stack inventory of Section 13.5, what the tool costs and includes today. The register of volatile facts in this book's bibliography lists the ten tables for review every quarter.

13.4 The Discovery Integration Map

A stack is a set of tools connected to a hub, and its quality is measured by its breaks. Figure 13.1 draws the discovery stack around its hub — the validation log and the contact base with consent and source — and the tools that feed it: the scheduling link, the forms and surveys, the landing pages, the payment tool, the whiteboard and the spreadsheet, and the AI assistant, which receives only de-identified notes and returns drafts.

Figure 13.1 — The discovery integration map. At the centre, the hub: the validation log and the contact base with consent and source — La Machine, or a spreadsheet and a CRM. Around it, the tools of discovery, each sending its data to the hub; the workflows of Chapter 11 carry the flows; the dashed arrow to the AI assistant carries only de-identified notes. A break — a manual export, a second list, notes in a personal drive, a page whose sign-ups go nowhere — is where evidence is lost. Original figure.

13.4.1 One hub

The hub is chosen once, in the first week, and everything connects to it. For a founder who will run the venture's marketing and sales in La Machine afterwards, the hub is the platform's contact base, with the log as a document on the project; for a founder who is not yet on the platform, it is a spreadsheet for the log and a free-tier CRM or a second spreadsheet for the contacts. What matters is not the tool but the rule: every observation and every person is in the hub, and every other tool writes to it or reads from it.

13.4.2 Finding the breaks

A break is a place where data must be moved by hand or is not moved at all, and a founder finds them by following one interviewee through the stack: the booking (did it create a contact?), the consent statement (was it sent and recorded?), the notes (are they on the profile, or on a laptop?), the survey response (did it reach the log?), the page sign-up (did it become a contact with a source?), the deposit (is it on the ledger, with the receipt's disclosures?), the summary (was it made from coded notes?). Each "no" is a break, and each break is fixed by a connection — a workflow of Chapter 11, a native integration, an automation tool from Table 13.10 — or by removing the tool that caused it. A stack with no breaks is one in which the weekly count of Chapter 11 needs no export.

13.5 The Total Cost of the Discovery Stack

The stack's cost is small in dollars and real in hours, and both are counted in a stack inventory that the venture keeps from the first day and that Book B1's marketing system and Book A2's plan will inherit.

13.5.1 Dollars and hours

The inventory lists, for every tool in use, its function, its owner (the account holder), its monthly cost and term, the data it holds and where, its connection to the hub, and the hours per month it consumes. The Do the Math box computes the quarter's total for a typical solo founder. Read the hours column as carefully as the dollars: a tool that costs nothing and consumes two hours a week of copying and pasting is the most expensive item in the inventory.

DO THE MATH The discovery stack for a quarter Illustrative figures. A solo founder's stack, chosen by the rules of Section 13.2. Free or already paid: the office suite's forms, spreadsheet and calendar booking page; a wall and sticky notes; Statistics Canada, Google Trends and the registers; the notes application already in use; the platform subscription if the founder is on La Machine — $0 additional. Paid at the scale of a test: a one-page site tool on a low annual tier for the landing page, about $25 for the quarter; a freemium whiteboard's paid tier for one month to export the boards, about $10; an automation tool's lowest paid tier for one month when a form needs to reach the hub, about $30; payment fees on four deposits of $150, about $20; an assistant's paid tier for the quarter, about $75 — a total of roughly $160, and under $300 with a scheduler's paid tier and a transcription tool for a month. Hours: about one hour a week before the workflows on exports, copies and reminders, and about fifteen minutes after them — the difference being the forty hours of Chapter 11's Do the Math box. The stack is not where the discovery quarter's money goes; the professionals of Chapter 12 and the founder's own hours are. A founder who has spent more than a few hundred dollars on tools before the gate has bought for a volume that has not arrived.

13.5.2 The stack inventory

The inventory is a table with one row per tool and eight columns — tool, function, owner, monthly cost and term, data held and where, connection to the hub, hours per month, review date — kept in the project and read at the monthly model review. It is the document that answers the four questions of Section 13.2.2 for every tool, and it is what Book B1's Chapter 13 will extend into the marketing stack when the venture reaches it.

QUÉBEC & CANADA NOTE Where the data lives (Québec and Canada, as of September 2026) Personal information collected in discovery — names, contact details, interview notes, responses, recordings — is subject to Québec's Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25) and, federally, to the Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5. Three consequences apply to the choice of tools. Purposes and consent: information collected for research is used for research, and a tool that reuses it for its own purposes must be disclosed and consented to (ss. 8, 12, 14). Communication outside Québec: before communicating personal information outside the province — which storing it with a cloud tool hosted elsewhere is — the enterprise must conduct a privacy impact assessment that considers the sensitivity of the information, the purposes, the protection measures and the legal regime of the destination, and may proceed only if the assessment shows adequate protection (s. 17); for a founder, the practical minimum is to know where each tool stores its data, to read its terms on retention and reuse, to record the assessment for the tools that hold personal information, and to prefer tools that document their locations and offer deletion on request. Governance: a person in charge of the protection of personal information, published policies and a register of incidents apply as soon as the enterprise holds personal information (ss. 3.1, 3.2). Book H2 (Privacy and Personal Information) gives the regime; Book K4 the rules for AI tools; the stack inventory's "data held and where" column is where the founder's assessment lives.

13.6 La Machine in the Stack

La Machine covers a specific set of the functions above, and the honest map of what it does and does not do is the last section of this chapter.

13.6.1 What La Machine covers

In the discovery quarter, the platform provides: projects, tasks with mandatory verification criteria, runs and the registry (Chapter 11); the registry of decisions (Chapters 1 and 10); contacts, companies, segments with live counts, the consent register and the do-not-reach list — the hub of Section 13.4 for a founder who will run the venture in the platform (Chapters 3, 5, 8); lead magnets whose consent box documents the consent at capture, forms, and landing pages served only when published, with responses and visits counted (Chapters 3, 9); sequences, the one path by which the module sends, whose messages are checked against the exclusion list, the identification and the consent flag before they leave (Chapters 9, 11; → see Book N5); meetings, calls and quotes for the sessions and the pilot proposals (Chapters 3, 9); the Business Explorer for a list of businesses and competitors in a region, within each source's cap (Chapters 5, 8); Calculator & Tools for the arithmetic of Chapter 7; the assistant, the AI agents and the orchestration working on the venture's own data under the approval rule (Chapter 12); Links and Notes for sources and the idea log (Chapter 2); and Books, where this library is read.

13.6.2 What it does not cover, and how it connects

The platform has, as of September 2026, no graphical canvas for the boards of Chapters 4, 6 and 10; no recording or transcription; no split testing on landing pages and no analytics beyond visits and responses; no market, census or trend data; no financial model beyond the calculators; no payment collection for deposits on a test page; and no directory of mentors, advisors or organizations — each of which the tables above cover with other tools. The Organizations page reads a model built for future multi-tenant accounts and does not describe the venture's own structure. The platform connects to the rest of the stack through the few exports its pages offer — the Business Explorer exports its list as a file, while the Contacts list and the pages of tasks, runs and projects export nothing — and through the integration options that Collection N describes for the venture's own workflows; a founder who is not yet on the platform loses nothing of the method — the hub is a spreadsheet and a CRM, and the workflows are built with the automation tools of Table 13.10 — and gains, on joining, a hub in which the contacts, the consent, the pages and the tasks already speak the format of this book.

WHO DOES THIS The stack in the three situations In a solo venture the founder chooses the hub in week one, keeps the inventory, and resists every tool that no task requires; the five tools of day one are enough for a month. In a founding team, one member owns the stack and the inventory, every account is in the venture's name with both founders as administrators, and the team agrees that a second list anywhere is a bug to be fixed the day it is found. In an established business, the intrapreneur keeps the new offer's evidence in its own hub — a project, a contact segment, a log — separate from the company's systems until the gate, so that the company's habits do not absorb the new model's data and so that a stop is clean; the company's tools are used where they cover a function and inventoried like any other. In all three, the "data held and where" column is filled for every tool that holds personal information, and the review date is kept.

Costly Mistakes

1. Buying for a volume that has not arrived. A CRM's paid hub, a page tool's growth plan and a transcription subscription in week one, for twelve interviews and one page. The consequence is a few hundred dollars a month for functions that a free tier or an existing tool covered, and a founder who feels the venture is "set up" when it has not met a customer.
2. Starting from tools instead of functions. A survey tool makes every question a survey; a page builder makes every test a page. The consequence is a method bent to the tools' idea of the work, and a founder who runs a survey where twelve interviews were needed.
3. Two lists. Sign-ups on a page, responses in a form, interviewees in a notes application, contacts in a CRM — none connected. The consequence is a weekly count that requires an hour of exports, a consent state nobody can find, and evidence lost at every break.
4. Notes on one laptop. The log in a personal notes application, the recordings in a personal cloud account, the boards in a personal whiteboard workspace. The consequence is a venture whose only asset lives outside its control — and, when a co-founder joins or a professional leaves, a scramble for access.
5. Personal information pasted into a tool nobody was told about. Twelve interviewees' words in a third-party assistant for a summary, without codes, without reading the terms, without the assessment. The consequence is a breach of the purpose and communication rules, discovered by the interviewee who asks what happened to their words.
6. Accounts in a professional's name. The page under the designer's account, the workflows under the builder's, the domain under an agency's. The consequence is a venture that cannot edit its own page after the engagement, or must pay to keep it — the mistake Chapter 12 named, made through the stack.

Key Takeaways

• Ten functions cover the discovery quarter — canvases, interview notes, surveys and forms, market data, landing pages, no-code prototypes, financial calculators, the log and contacts, scheduling and payments, AI assistants and automation — and each is specified by what it must do before any tool is named.
• Four rules keep the stack small: free and already-paid first, connect everything, one source of truth, know where the data lives; and four questions precede any purchase — which task requires it, what it replaces, who owns the account, what happens to the data when you leave.
• The tables give pricing models, not amounts, as of September 2026, because amounts change several times a year and the library cites a price only from a dated vendor page; La Machine appears in the functions it covers, described in the same terms as the alternatives.
• Five tools cover day one — a log, a form with unbundled consent, a scheduling link, a spreadsheet for the numbers, a contact base with consent and source — most of them free or already paid; a page tool is added when the test comes.
• The stack is drawn around one hub — the log and the contact base — and its quality is measured by its breaks, found by following one interviewee through every tool.
• The stack costs a few hundred dollars for the quarter and about an hour a week before the workflows; the inventory — tool, function, owner, cost, data and where, connection, hours, review date — is kept from the first day and inherited by the marketing stack and the plan.
• Personal information stays where the rules allow: purposes and consent respected, a privacy impact assessment before it is stored outside Québec, a person in charge and policies as soon as the venture holds it.
• La Machine covers the hub, the tasks, the decisions, the forms, lead magnets and pages, the sequences, the sessions and quotes, the Explorer, the calculators and the assistant — and says plainly what it does not: a canvas, transcription, split tests, market data, a financial model, deposits on a test page, a directory.

Your Action Plan

The plan below takes about three hours in the first week and one hour at each monthly review, and it installs the stack that Chapter 11's workflows run on. Steps 1 and 2 choose the hub and the five tools; step 3 draws the map; step 4 fills the inventory; step 5 answers the data question; step 6 sets the review.

In La Machine

La Machine is the hub of Figure 13.1 for a founder who will run the venture in the platform, and its Tools page is the inventory's model.

• The Tools page (/outils) is the platform's own inventory of what is installed and connected to make it work, and its rule is the one this chapter adopts: an inventory is measured, never declared — version, location and date for every tool, and a hosted service simply has no version to show; a founder's stack inventory of Section 13.5.2 follows the same discipline, with the "data held and where" column as its measured field.
• Calculator & Tools (productivity/tools) provides the calculators of Table 13.7; contacts, companies, segments, consent and do-not-reach are the hub of Table 13.8; forms, lead magnets and landing pages cover Tables 13.3 and 13.5 with the approval rule; meetings, calls and quotes cover the scheduling side of Table 13.9; the Business Explorer covers the directory rows of Table 13.4; the assistant, the AI agents, orchestration and the registry cover Table 13.10 under the approval rule; Links and Notes hold the sources and the idea log.
• Books (business/assets/books) is where this library is read inside the platform — this book, Book B1 for the marketing stack this inventory grows into, and Collection N for the platform page by page.

What La Machine does not do is listed in Section 13.6.2 and repeated here for the founder who wants the short version: no canvas, no transcription, no split tests, no market data, no financial model, no deposits on a test page, no directory. Book N1 (Getting Started with La Machine) walks through the navigation; Book N3 (Workflows, Automation and Human Approval) shows how the workflows of Chapter 11 are built in the process engine that the platform opens through a page, registered and connected to the tools of this chapter; Book K1 (The SME Technology Stack) develops the selection rules of Section 13.2 for the whole business.

Self-Check

1. Name the ten functions of the discovery stack and explain why the chapter specifies functions before naming tools.
2. State the four selection rules and the four questions to ask before any purchase.
3. Why does the chapter print pricing models rather than amounts, and how should a founder read the tables a year from now?
4. What is the hub of the integration map, and how does a founder find the breaks in a stack?
5. What three consequences does Québec's privacy law have for the choice of tools that hold personal information?

Answers. (1) Canvases and whiteboards; interview notes, recording and transcription; surveys and forms; market, trend and competitor data; landing pages and smoke tests; no-code prototypes and MVPs; financial modeling and calculators; the validation log, contacts and CRM; scheduling and payments; AI assistants and workflow automation. A founder who starts from a tool inherits the tool's idea of the work; one who starts from the function chooses the cheapest tool that does what the method requires (Section 13.1). (2) Free and already-paid first; connect everything; one source of truth; know where the data lives. Which task requires it; what it replaces; who owns the account; what happens to the data when we leave (Section 13.2). (3) Amounts change several times a year and the library cites a price only from a dated vendor page; the models change less often; the tables are a map of the function and a starting list of names, to be verified on the vendor's page with the date recorded in the inventory, and reviewed quarterly through the register of volatile facts (Section 13.3, What Changed). (4) The validation log and the contact base with consent and source; by following one interviewee through every tool — booking, consent, notes, response, sign-up, deposit, summary — and marking each "no" as a break to connect or a tool to remove (Section 13.4). (5) Purposes and consent respected, with a tool's reuse disclosed; a privacy impact assessment before personal information is communicated outside Québec, which cloud storage elsewhere is (P-39.1 s. 17); and governance — a person in charge, policies, an incident register — as soon as the venture holds personal information (Section 13.5, Québec and Canada Note).

Chapter References

Act respecting the protection of personal information in the private sector, CQLR c P-39.1 (as reformed by Law 25), ss. 3.1, 3.2, 8, 12, 14, 17. LégisQuébec. Retrieved September 7, 2026, from https://www.legisquebec.gouv.qc.ca/en/document/cs/P-39.1

Personal Information Protection and Electronic Documents Act, S.C. 2000, c. 5. Justice Laws Website. Retrieved September 7, 2026, from https://laws-lois.justice.gc.ca/eng/acts/P-8.6/

Business Development Bank of Canada. (2026c). Business plan template and tools. Retrieved September 7, 2026, from https://www.bdc.ca/en/articles-tools/business-strategy-planning/define-strategy/business-plan-template

Commission d'accès à l'information du Québec. (n.d.). Entreprises : protection des renseignements personnels. Gouvernement du Québec. Retrieved September 7, 2026, from https://www.cai.gouv.qc.ca/

Statistics Canada. (2026b). Statistics Canada. Retrieved September 7, 2026, from https://www.statcan.gc.ca/

Strategyzer AG. (n.d.). The Business Model Canvas. Retrieved September 7, 2026, from https://www.strategyzer.com/library/the-business-model-canvas

Vendor pricing pages consulted for the pricing models of Tables 13.1 to 13.10 (Miro, FigJam, Strategyzer, Mural, Lucidspark, Otter.ai, Notion, Dovetail, Google Forms, Microsoft Forms, Typeform, SurveyMonkey, Tally, Jotform, Google Trends, Similarweb, Semrush, Carrd, Webflow, Framer, Unbounce, Figma, Bubble, Softr, Glide, Airtable, Shopify, Gumroad, LivePlan, HubSpot, Zoho, Pipedrive, Calendly, Stripe, Square, PayPal, the AI assistant vendors of Table 13.9, n8n, Zapier, Make). Models as observed September 7, 2026; no amounts cited; verify on each vendor's page before budgeting.

Chapter 14 — Practical Advice, Pitfalls and Your Action Plan

Thirteen chapters have given the founder a method, a calendar and a set of tools. This last chapter gives the advice that experience adds to method, the mistakes that end ventures before their first customer, the ninety-day plan that sequences everything on a calendar, the ten numbers that say whether it is working, the audit that keeps it honest, and the map of where to go next in this library. Read it before the quarter begins and again on day ninety.

What You'll Be Able to Do

• Apply fifteen pieces of advice that change the result of a discovery quarter, each tied to the chapter that explains it.
• Recognize the ten costly mistakes of early-stage ventures in your own plan before they cost you a year.
• Run the 30-60-90-day plan — decide, build, test and decide — with deliverables and verification at the end of each phase.
• Watch ten numbers, read them together, and know what each one is telling you.
• Run the one-hour discovery audit at day ninety, and use the master checklist that gathers the fourteen action plans.
• Choose the next book on the outcome of the gate and the number that is weakest.

Why It Matters

A method that is not sequenced is a reading; a plan that is not measured is a hope. The ninety-day plan turns the fourteen chapters into thirteen weeks that a solo founder can keep beside a job, and the ten numbers turn the quarter into something a co-founder, a mentor, a lender or the founder's own future self can judge without opinion. For the founder, this chapter is the difference between having read about discovery and having done it. For a team, it is the shared calendar and the shared numbers. For an established business, it is the mandate's schedule and the gate's evidence.

In This Chapter

• 14.1 Fifteen Pieces of Advice That Change Results
• 14.2 The Costly Mistakes of Early-Stage Ventures
• 14.3 Your 30-60-90-Day Plan
• 14.4 The Ten Numbers to Watch
• 14.5 The Master Checklist
• 14.6 Where to Go Next in the Library

14.1 Fifteen Pieces of Advice That Change Results

The advice below is what founders who have done a discovery quarter tell founders who are about to. Each piece is a complete instruction, and each names the chapter that gives it its mechanism.

1. Write the problem before the solution, and keep three. A founder who loves the problem uses every piece of evidence to find the solution that fits; a founder who loves the solution defends it against the evidence. Write the opportunity statement in the customer's words before any product is named, keep three briefs alive, and let the interviews choose (Chapter 2; Blank & Dorf, 2012).

2. Ask about the last time, never about the future. The only reliable information a conversation yields is about the customer's past behavior: the last time the problem happened, how often, what they did about it, what it cost. Compliments and predictions are free and worthless; the script that asks about the past and ends with a request for a commitment is the whole skill (Chapter 3; Fitzpatrick, 2013).

3. Set the threshold before the result, every time. A threshold set in advance turns a result into a decision; one set afterwards turns a decision into a rationalization. Write it on the card, in the log's header, on the test — and never move it after the number arrives (Chapters 3, 9, 10).

4. Count people, not statements. One vivid interviewee who said the same thing five times is one person. A pattern is six of twelve different people saying it unprompted; a signal is three to five; two is an anecdote. The synthesis is done on cards, and the cards are counted by person (Chapter 3).

5. Climb the ladder — says, does, pays — and decide at the rung the decision deserves. Choose a problem on what people say; choose a segment on what they do; build, quit a job or borrow on what they pay. Four deposits from strangers outweigh forty survey answers from friends (Chapters 3, 9).

6. Beat "doing nothing" before you beat anyone else. Your first competitor is the customer's tolerance of the problem. Make the cost of the status quo visible in the customer's own numbers before the price is mentioned, and rate the "do nothing" column of the alternatives grid honestly (Chapters 4, 8).

7. Lower the switching cost as much as you raise the gain. Customers weigh losses about three times more than gains, and developers overvalue their product's benefits about three times. A guarantee, a free first month or a simpler start often wins the switching test where a feature loses it (Chapter 4).

8. Size bottom-up, and turn the size into a weekly number. One percent of a huge market is a wish; a chain of counts, shares, prices and frequencies is a plan, and the plan becomes eight contacts a week. If the weekly number does not fit the hours, the target, the channel or the model changes now, not in month nine (Chapter 5).

9. Win one segment decisively before touching a second. A beachhead is chosen on money, urgency, deliverability, channel, competition, word of mouth, adjacency and fit — not on size — and it is won when its members refer each other without being asked. The runner-up is written down as the pivot (Chapter 5).

10. Fill the canvas from the right side, and mark every note. The customer side comes from evidence, the delivery side from asking what it takes, and every note is evidence or hypothesis, in two colors, versioned and dated. A canvas that is a poster is worthless; one that is a hypothesis board is the venture's instrument (Chapter 6; Osterwalder & Pigneur, 2010).

11. Include your own salary in the break-even. A model that works only because the founder is unpaid is a subsidy, not a model. The break-even triples when the imputed salary is added; compute it that way, compute the runway with and without the day job, and keep the taxes out of revenue (Chapter 7).

12. Test the assumption that kills you first, with the cheapest test that can change your mind. The killing assumption tested in week six costs six weeks; in month fourteen it costs the venture. Rank on importance and evidence; test the upper-left corner; climb the test ladder only as far as the next decision requires (Chapters 9, 10; Bland & Osterwalder, 2020).

13. Sign nothing personal before the riskiest assumptions have survived. A lease, a loan, a guarantee, a full build or a brand bought before the tests commit money to a model discovery will change — and make the founder unable to hear that it must (Chapters 1, 10, 12; Wasserman, 2012).

14. Keep the cadence, not the intensity. Six hours a week for thirteen weeks is seventy-eight hours and a decision; twenty hours for two weeks is forty hours and an abandoned venture. Book the ring before the quarter starts, file everything within the hour, and hold day ten even when nothing changed (Chapter 11).

15. Record every decision, and never edit the record. The question, the options, the evidence, the decision, the reasoning, the date and the trigger that would reopen it — written before the decision is announced, kept where a co-founder, a lender or the founder's future self can read it. The registry is the venture's memory, and a stop recorded on criteria with a full log is the cheapest outcome a discovery quarter can have (Chapter 10).

KEY CONCEPT Evidence compounds; opinions do not Twelve interviews, a page test, a concierge and two pilots produce, in a quarter, a body of evidence that no opinion can argue with and that every later document — the plan, the financing file, the pitch — draws on. The same quarter spent building produces a product and no evidence, and every later document starts from a guess. Evidence is the only asset of a venture before its first customer, and it compounds: each test narrows the next, each log entry makes the next synthesis faster, each decision record makes the next decision cheaper. The founder who finishes the quarter with a full log and a stop has more than the founder who finishes with a product and no customers.

14.2 The Costly Mistakes of Early-Stage Ventures

Table 14.1 gathers the ten mistakes that end ventures before their first customer, drawn from the costly-mistakes sections of the previous chapters. For each it gives the pattern as a founder would recognize it, the consequence in time and money, and the chapter whose method prevents it. Read it as a checklist against your own plan: a founder who finds two of these in the current draft has found the first two tasks of the quarter.

Table 14.1 — Ten costly mistakes of early-stage ventures

WATCH OUT Building before validating Of the ten, the first and the ninth are the same mistake at two moments, and together they are the leading cause in every post-mortem, where poor product-market fit is cited by 43% of the failed companies whose reasons could be identified (CB Insights, 2026; Chapter 1): the founder builds because building feels like progress and testing feels like doubt. The tell-tale signs are a development quote before a page test, a brand identity before a repeat test, an inventory before a pre-order, a lease before a pilot. The consequence is the untested year of Chapter 1 — about $84,000 in money and displaced income in the illustrative case — against a discovery quarter of about $7,900. The remedy is not to build less; it is to build after, when the riskiest assumptions have survived and the build is the next test rather than the first.

14.3 Your 30-60-90-Day Plan

The plan sequences the twenty setup tasks of Chapter 11 over thirteen weeks in three phases: days 1 to 30 — decide (the problem, the evidence, the people around the venture); days 31 to 60 — build (the proposition, the segment, the model, the numbers, the competitive position, the map); days 61 to 90 — test and decide (the riskiest assumptions on the ladder, the register, the gate). It assumes six hours a week for a solo founder — a team divides the phases, an established business staffs them — and a discovery budget of a few thousand dollars at most.

14.3.1 The shape of the plan

Table 14.2 lays the plan out in eight periods. For each it names the focus, the actions with their chapters, the deliverables, and the verification at the end of the period. Read the last column as the gate of each period: a period whose verification is not met is not finished, and the next period waits — which is what keeps the plan from becoming a list of things that were started.

Table 14.2 — The 30-60-90-day plan (solo founder on six hours a week; a team divides, an established business staffs)

14.3.2 Reading the plan honestly

Three warnings. The plan is the discovery, not the launch. By day ninety the founder has a validated model, a refuted one with a pivot, or a stop with a full log; the product, the launch and the customers beyond the pilots belong to the next books. The hours are real. Six hours a week for thirteen weeks is seventy-eight hours, plus the accountant's four, the lawyer's two, the designer's two days and the mentor's meetings; a plan that assumes more hours than the founder has will be abandoned in week five — cut scope, not cadence. Compliance is not phase three. The consent statement, the claims file, the disclosures on the page and the pre-order rules are in place before the first interview, the first page and the first deposit, because nothing lawful can be sent, published or collected without them, and the fines are real.

DO THE MATH What ninety days of discovery is worth Illustrative figures, consistent with the worked examples of the earlier chapters. The quarter costs a solo founder about seventy-eight hours ($5,850 at $75), the professionals of Chapter 12 (about $3,200), the tools of Chapter 13 (about $300) and the deposits' fees — about $9,400 including the founder's time, or about $3,500 in cash. What it buys is measured three ways. The year prevented: in the Rideline case of Chapter 1, fourteen months and $62,000 spent on a job that twelve interviews would have found in month one — the quarter's value there is the whole year. The model corrected: in Mariam's case, the interviews moved the segment, killed the shared-folder feature and doubled the price the value map could support; the accountant's first hour moved the break-even from seven clients to twenty-five, which changed the first-year target and the hiring plan; Norvale's route arithmetic moved the beachhead from the city to two streets. Each correction, made on paper, would have cost a quarter or a year made in the market. The financing file: a founder who arrives at a lender or a program with a log, a dated map, three commitments and a decision record is read as evidence; one who arrives with a plan built on a share of a billion is not — and the difference is a loan, a program's mentoring, a co-founder's two years. The quarter does not guarantee the venture succeeds; it guarantees that the next dollar is spent on a model whose riskiest assumptions have met customers.

CASE IN POINT Ninety days from idea to validated model Fictional example; names and figures are invented; the case gathers the receipt-clinic story of Chapters 2 to 13. Mariam Haddad started on the first Monday of a quarter with an idea for a receipt-scanning app and six hours a week. Days 1 to 30. The nine-answer sheet marked seven of nine answers as guesses; the idea log produced twenty-three problems in two weeks, and the scorecard ranked "the two weeks before the tax deadline" first; three briefs were written and the first chosen; twelve owners were recruited, six of them strangers reached through the association; six interviews and one synthesis killed the shared-folder feature and split the segment at about $300,000 of revenue; the accountant's first hour asked about her own salary. Days 31 to 60. Round two confirmed nine of twelve on the March pain, five said "ashamed" unprompted, seven asked when the clinic would start and four paid a $150 deposit; the profile, the map and the statement passed the repeat test with four of five strangers in the "quarterly bookkeeping session" frame; the sizing put the regional SAM at about 4,400 businesses and the capacity-bounded SOM at eighty clients by year three; the beachhead was the association's members above $300,000 with staff, the runner-up the referring accountants' clients; the canvas was drawn three ways and the service chosen for year one with the licensed kit as the next pin; the numbers gave $85 of contribution a session, a $150 acquisition cost, a five-month payback, a break-even of twenty-five clients with her salary included, and three numbers that must be true — the price above $150, the renewal above 60%, eight contacts a week; the alternatives grid rated "do nothing" honestly and confirmed the "only": the sorting done with the owner at a fixed fee. Days 61 to 90. A page test through the association's newsletter converted 9% of fitting visitors at $180 against a threshold of 8%; a concierge of six paid sessions logged two hours and ten minutes per session and produced the envelope kit as the thing clients kept; two pilots with agreements ran through the quarter's end; the register recorded her exposure — no lease, no loan, a second bookkeeper as a step cost at forty clients; the pre-mortem before the second bookkeeper's hire listed "the association stops publishing" and added a second channel. At the gate, the ten numbers were written, the audit found two fixes, and the decision record said persevere: the price and the weekly contacts were validated, the renewal rate a proxy until March, the next book A4 for the legal form and the registration, then A2 for the plan the accountant asked for. Nothing spectacular had happened. A model that had begun as an app nobody wanted had become a service twelve people had paid for, on evidence a lender could read.

14.4 The Ten Numbers to Watch

A discovery quarter is judged on ten numbers, read together at the sprint reviews and the gate. Table 14.3 lists them with what each one tells the founder and the chapter that produces it; the sixth, the share of pilot customers who would be very disappointed to lose the offer, is the practitioners' test of a must-have (Ellis & Brown, 2017). Read them as a set: a strong pattern with no commitments says the problem is real and the offer is not; commitments with a weak contribution say the offer sells and the model does not; a good weekly number with a short runway says the venture is right and the founder cannot afford to wait for it.

Table 14.3 — The ten numbers of the discovery quarter

14.4.1 Reading them together

Four readings recur. Numbers 1 to 3 strong, 4 and 5 weak: the problem is real and the proposition, the price or the frame is wrong — back to Chapter 4 with the page's words. Numbers 4 and 5 strong, 7 weak: the offer sells and the model does not — Chapter 7's stress test, then a value-capture or a business-architecture pivot. Numbers 1 to 8 strong, 9 weak: everything works and the channel cannot deliver the volume — an engine-of-growth or a channel pivot (Ries, 2011), or a smaller first-year target. Everything strong, 10 short: the venture is right and the founder cannot afford to wait — the financing of Book A5, or the day job kept longer. And one reading that ends the quarter: numbers 2 and 3 weak after twelve strangers and a second brief exhausted — the stop of Chapter 10, on criteria, with the log.

14.5 The Master Checklist

The fourteen action plans of this book contain about eighty verification criteria, and the companion file A1-Checklist gathers them all, chapter by chapter, for the founder to mark yes, partly or no. This section gives the one-hour audit that reads the checklist in order at day ninety — and again at any later gate — and produces at most three fixes.

METHOD The one-hour discovery audit When to use it: at day ninety before the gate, and at any later review of the model. Steps: (1) 0–10 minutes — the cadence. Were the thirteen slots held, the sprint reviews on day ten, the model reviews on the first Monday? If two were skipped, stop: the first fix is the cadence, and the rest of the audit reads a quarter that did not happen. (2) 10–25 minutes — the evidence and the rules. Twelve interviews, half strangers, notes within the hour; the consent statement read every time; the claims file complete; the page and the deposits within the rules; recordings deleted; no research contact on a marketing list. (3) 25–45 minutes — the model and the numbers. The brief versioned; the profile from patterns; the statement passed the repeat test; the sizing computed twice; the beachhead chosen on the eight criteria with a runner-up; the canvas marked and dated; the break-even with the imputed salary; the three numbers that must be true; the alternatives grid with its "do nothing" column; the advantage sentence honest. (4) 45–55 minutes — the tests and the decisions. Every test card had a threshold set before; every result was read with an outsider; the map redrawn after every test; the register has the founder's row; the pre-mortem was run before every commitment; every decision is in the registry with its seven parts. (5) 55–60 minutes — decide. Three fixes at most, each with an owner, a criterion and a date, entered as tasks. Deliverable: the marked checklist and three tasks. Pitfalls: auditing the results instead of the method — the audit asks whether the evidence was collected and read correctly, not whether it was favorable; and skipping section 1, which is where most quarters actually failed.

14.6 Where to Go Next in the Library

The gate's outcome and the weakest of the ten numbers decide the next book. Table 14.4 maps the outcomes to the books of this library, with the reason each one comes next. Read the first three rows as the three outcomes of the gate; the rest as the specific weakness the numbers revealed. Most founders who reach a go decision read Book A4 and Book A2 next, in that order, because the accountant's and the lender's questions arrive together.

Table 14.4 — Where to go next