Entrepreneurship & New Ventures

The big picture

Walk down one commercial street in any American town and you will pass three people the newspapers would all call entrepreneurs. The first owns the bakery on the corner. She has four employees, buys flour by the pallet, works six days a week, and takes home somewhere in the range of sixty thousand dollars in a good year. The second runs a plumbing company out of a truck yard behind the hardware store. He started alone twelve years ago and now dispatches eight vans, employs eleven people, and clears a good deal more than the baker. The third rents a desk above the pharmacy and is building software for dental practices. She has no revenue, two cofounders, and eight months of savings.

These three people are doing genuinely different jobs. Their risks are different, their financing is different, their definitions of success are different, and advice that helps one of them can wreck another. Yet the word entrepreneur is applied to all three, and most of the advice in circulation is written for the third one, who is by far the rarest. Sorting out this confusion is the first job of this course, because almost every bad decision students make in their first venture traces back to importing a playbook written for a different kind of business.

One honest note at the start, and it will recur. A text course can give you frameworks, real numbers, worked arithmetic, and the evidence for and against common claims. It cannot give you a customer who says no to your face, and it cannot substitute for a licensed attorney or accountant when your specific situation calls for one. Treat this as preparation, not permission.

Four old definitions that still do work

The word entrepreneur arrived in economics through Richard Cantillon, an Irish-French banker writing in the 1730s. Cantillon noticed a class of people who bought at known prices and sold at unknown ones: the farmer who plants at a known cost and sells at whatever the harvest market pays. His entrepreneur is defined not by innovation or by ambition but by bearing uncertainty. That is still the cleanest core of the idea, and it already applies to all three people on our street.

Joseph Schumpeter, writing between 1911 and the 1940s, added something sharper. For Schumpeter the entrepreneur carries out new combinations: a new good, a new method of production, a new market, a new source of supply, a new form of organization. Growth in his account does not come from tidy accumulation but from creative destruction, in which new combinations displace the old ones and destroy their value. Schumpeter is the reason we associate the word with disruption. He is also the source of a common error, because on his strict definition the plumber who copies an existing business model is not an entrepreneur at all. He is a business owner. That distinction is useful analytically and useless morally, and this course keeps the analytic part while discarding any implication that copying a working model is a lesser activity.

Frank Knight, in 1921, drew the line that matters most for how you think about risk. Risk is a situation with a known distribution of outcomes: a fire in a warehouse has an actuarial probability, which is why it can be insured. Uncertainty is a situation where you cannot even write down the distribution, because the event is genuinely novel. Knight argued that entrepreneurial profit is payment for bearing uncertainty specifically, because anything insurable would simply be priced into a cost. This is why the honest answer to "what are my chances?" is usually a base rate plus a shrug, and why the second lesson in this course spends its time on base rates.

Israel Kirzner, working in the Austrian tradition decades later, supplied the counterweight to Schumpeter. Kirzner's entrepreneur is not a heroic disruptor but an alert person who notices a mismatch others walked past: a price difference, an unmet need, a resource sitting idle. On this account the woman who realizes her town of thirty thousand has no decent daycare and opens one has done exactly what entrepreneurship is for. Nothing was invented. Something was noticed and acted upon.

Key idea: Bearing uncertainty is the common core; innovation, alertness, and organization-building are the different things entrepreneurs do with it. For this course, entrepreneurship means creating a new organization under uncertainty while bearing the residual outcome, good or bad.

Six different animals wearing one word

Here is a taxonomy worth memorizing, because you will spend the rest of the course asking which row you are in.

Type	Example	Typical funding	Success looks like
Self-employed / nonemployer	Freelance editor, sole plumber, Etsy seller	Savings, a credit card, revenue	A steady personal income, control of schedule
Small business owner (employer firm)	Bakery, dental practice, eight-van plumbing firm	Savings, bank or SBA loan, retained profit	Profitable operation, a salable asset, jobs created
Scalable startup	Software, biotech, marketplaces	Angel and venture capital, sometimes never profitable early	Large market share, acquisition or public offering
Franchisee	Subway, a tax-prep office	Franchise fee plus loan, often SBA	Operating a proven model within a system
Social venture / nonprofit	Community clinic, workforce training program	Grants, donations, earned revenue	Measured impact plus financial durability
Acquirer (entrepreneurship through acquisition)	Buying a retiring owner's HVAC company	Seller financing, SBA 7(a), search fund investors	Owning existing cash flow and improving it

Corporate entrepreneurship, sometimes called intrapreneurship, is a seventh case: building something new inside an existing company, with the firm's resources and the firm's politics. It is a real career path, and the skills in this course transfer to it almost entirely, minus the part where you personally guarantee a lease.

Key idea: The six types differ in funding, risk profile, time horizon, and exit. Deciding which one you are building is not a branding question; it determines which advice applies to you.

What the American business population actually looks like

Now the numbers, because intuitions here are badly calibrated. The U.S. Small Business Administration's Office of Advocacy counts roughly 33 million small businesses in the United States, which is about 99.9 percent of all American firms. That headline hides the important structure. Of those 33 million, only about 6 million are employer firms, meaning they have at least one employee on payroll. The other roughly 27 million are nonemployer businesses: one person, no payroll, often part time. Roughly four out of five American businesses have no employees at all.

Go one level deeper into the 6 million employer firms and the distribution stays lopsided. The large majority have fewer than 20 employees, and a substantial share have fewer than five. At the other end, firms with 500 or more employees number in the tens of thousands, not the millions. Small firms in aggregate employ close to 46 percent of American private-sector workers, which means the other half of the workforce is at a comparatively tiny number of large companies.

Flow matters as much as stock. The Census Bureau's Business Formation Statistics counts business applications, essentially requests for a federal employer identification number. In recent years that has run around five million applications annually, a level that jumped sharply in 2020 and has stayed high. Only about a third of those are classified as high-propensity applications, meaning they show characteristics associated with actually becoming employer businesses. Many of the rest are side projects, single consulting shingles, and applications that never turn into anything.

Against those five million applications, set this: the number of American companies that receive a first round of venture capital in a year is a few thousand. Not a few hundred thousand. A few thousand. Venture capital touches something on the order of one in a thousand new business applications, and a far smaller fraction of the businesses that actually operate. That single ratio explains why so much popular entrepreneurship advice does not fit the person reading it.

Key idea: The modal American business is one person with no employees. The venture-backed startup is a statistically tiny corner of entrepreneurship that occupies most of the coverage.

Why the mismatch does real damage

Consider our baker reading a popular startup book. She learns to prioritize growth over profitability, to raise money before revenue, to treat the first two years as a search for product-market fit funded by investors, and to expand into new cities quickly to claim the market. Every one of those instructions is defensible for a software company with near-zero marginal cost and a plausible path to national scale. Applied to a bakery, they are close to a suicide note. Her marginal cost is flour and labor, her market is a two-mile radius, there is no investor for a single bakery, and a second location opened before the first one throws off reliable cash is the classic way that successful bakeries die.

The reverse error is just as real. A founder building a diagnostic device that needs three years of engineering and a regulatory clearance before its first legal sale cannot be profitable in month one, and telling her to bootstrap on revenue is telling her not to build it. Different production functions demand different financing.

So take the ordinary business seriously, because it is what most people reading this will actually start. Run the plumber's numbers. Eight vans, each producing roughly three hundred thousand dollars of annual revenue, gives about 2.4 million dollars of revenue. At a net margin of twelve percent after paying himself a modest salary, that is roughly 290 thousand dollars of owner earnings. Businesses of that size commonly change hands at somewhere around two to three times seller's discretionary earnings, so the firm he built out of one truck is plausibly a six or seven hundred thousand dollar asset on top of the income it pays him. Nobody wrote a magazine profile about it. It is, by any sane standard, a successful venture.

Key idea: Match the playbook to the business. Growth-at-all-costs advice assumes near-zero marginal cost, a large addressable market, and access to outside capital, and it is destructive when any of those assumptions fails.

Try it

Classify each of the following, then say what the most dangerous piece of borrowed advice would be for it. (1) A former nurse launching a home health agency with six aides. (2) Two graduate students building an AI tool for radiology that will need clinical validation. (3) A carpenter who quits to take custom furniture commissions alone. (4) A regional manager who buys a retiring owner's commercial landscaping firm with an SBA loan.

Answer: (1) Small business employer firm; the dangerous advice is to scale headcount before payroll cash flow and credentialing can support it. (2) Scalable startup; the dangerous advice is to launch fast and iterate on live users, which regulation and patient safety forbid. (3) Self-employed nonemployer; the dangerous advice is to raise outside capital, since there is no investor return in a one-person craft business and debt would only add fixed cost. (4) Entrepreneurship through acquisition; the dangerous advice is to change everything in month one, since the value bought was the existing customer and crew relationships.

Common misconceptions

• "Entrepreneur means startup founder." It covers everyone bearing the residual risk of a new venture, and the overwhelming majority of them are running ordinary local businesses.
• "Real entrepreneurs innovate; copying is not entrepreneurship." On Schumpeter's strict definition, imitation is not entrepreneurship, but a well-run copy of a proven model in an underserved place creates real value and real income.
• "Small business is a small part of the economy." Small firms employ close to half of American private-sector workers.
• "Most businesses have employees." Roughly four in five American businesses are nonemployer firms.
• "You need venture capital to start something serious." Venture capital reaches a few thousand companies a year out of roughly five million applications, and it is inappropriate for most business models.
• "Risk and uncertainty are the same thing." Knight's distinction is load-bearing: insurable risk gets priced, while genuine uncertainty is what entrepreneurial profit compensates.

Recap

• Bearing uncertainty is the shared core; Schumpeter adds new combinations, Knight adds the risk-uncertainty distinction, Kirzner adds alertness to unnoticed mismatches.
• Six kinds of venture wear the same word: self-employed, small business employer, scalable startup, franchisee, social venture, and acquirer, plus corporate entrepreneurship inside firms.
• The United States has roughly 33 million small businesses, of which about 27 million have no employees and about 6 million are employer firms.
• Recent years have seen roughly five million business applications annually, while only a few thousand companies a year receive first venture funding.
• Advice is not universal: growth-first playbooks assume low marginal cost, large markets, and outside capital, and they damage businesses that lack those features.
• An eight-van plumbing firm can produce a quarter-million dollars of annual owner earnings and a salable asset, which is a serious entrepreneurial outcome even though no one profiles it.

Sources

1. U.S. Small Business Administration, Office of Advocacy. (2024). Frequently asked questions about small business. advocacy.sba.gov
2. U.S. Census Bureau. (2025). Business Formation Statistics. census.gov
3. Shepherd, D. A., et al. (2020). Entrepreneurship today. In Entrepreneurship. OpenStax, Rice University. openstax.org
4. Encyclopaedia Britannica. (2025). Entrepreneurship. britannica.com
5. Wikipedia contributors. (2025). Creative destruction. en.wikipedia.org

The big picture

You have heard that ninety percent of startups fail. You have probably also heard that most businesses fail in the first year. Both claims circulate endlessly, neither has a rigorous source, and the actual data are both better and more interesting than the folklore. This lesson is the honest floor of the course. Everything after it, the customer interviews, the pricing arithmetic, the cap table, is an attempt to move your personal odds within a distribution whose shape you should know before you start.

There is a second and larger problem with what you have heard, and it is not that the numbers are wrong. It is that the stories are drawn from a biased sample. You know the founder who dropped out and built a company worth billions. You do not know the tens of thousands who dropped out and are now thirty-four with no degree and no company, because nobody writes those books, gives those talks, or funds those documentaries. Correcting for that distortion is a skill, it has a name, and once you have it you will notice it in almost every piece of business advice you encounter for the rest of your life. Let us install it early.

What the survival data actually say

The U.S. Bureau of Labor Statistics runs a program called Business Employment Dynamics, which tracks establishments quarterly using state unemployment insurance records. It is close to a census rather than a survey, which makes it the best survival evidence available. Track a cohort of establishments born in a given year and the pattern is strikingly consistent across decades:

Years since opening	Approximate share still operating
1 year	Around 80 percent
2 years	Around 70 percent
5 years	Around 50 percent
10 years	Around 35 percent
15 years	Around 25 to 30 percent

Read the table honestly in both directions. Roughly one in five new establishments is gone within twelve months, which is worse than optimists expect. But half are still operating at five years, which is far better than the folklore of near-certain doom. And the ten-year figure is the one to hold onto: most new businesses do not survive ten years, and that has been true through booms, recessions, and technology waves. It is a structural feature of an economy where entry is easy, not a verdict on any individual.

Three caveats keep you honest with these numbers. First, BLS tracks establishments, meaning physical locations, not firms. A company that closes one location and opens another registers a death and a birth. Second, and more important, a closure is not a failure. Owners retire. Owners sell. Owners get a job offer they cannot refuse, or their spouse gets transferred, or the lease ends and they decide they are tired. Research on business exits consistently finds that a substantial share of closures involve businesses the owner considered successful. The data have a column for closure and no column for regret. Third, survival is not success. A business can survive fifteen years while paying its owner less than a job would.

Key idea: Roughly 80 percent of new establishments reach year one, about half reach year five, and about a third reach year ten. Most new businesses do not survive a decade, and many closures are exits rather than failures.

The venture numbers, which are worse and differently shaped

Venture-backed startups are a different distribution entirely, and it is worth seeing early so that Module 4 makes sense. An often-cited analysis by Correlation Ventures covering roughly 21,000 financings from 2004 to 2013 found that about 65 percent of venture investments returned less than the capital invested, while roughly four percent returned ten times or more. That is a power law: the median outcome is a loss, and nearly all of the fund's return comes from a handful of investments.

Notice who that distribution is designed for. A venture fund holding thirty investments can survive twenty-five failures if one returns fifty times its money. A founder holds exactly one investment. The fund is diversified across the power law; you are a single draw from it. Investor risk tolerance and founder risk exposure are not the same thing, and a great deal of bad advice comes from an investor generalizing their portfolio experience to your single life.

Key idea: Venture returns follow a power law in which the median deal loses money and a few outliers carry everything. A diversified fund can live with that shape. A founder with one company cannot diversify.

Survivorship bias, named and understood

Survivorship bias is the error of drawing conclusions from a sample that has been filtered by success, while treating it as if it were a sample of everyone who tried. It is the single most consequential reasoning error in entrepreneurship, and the standard illustration comes from wartime statistics.

During the Second World War, the Statistical Research Group at Columbia University, which included the mathematician Abraham Wald, was asked to advise on where to add armor to aircraft. Armor is heavy, so it can only go in a few places. The available evidence was the damage pattern on planes that came back from missions. Wald's contribution, set out in a series of technical memoranda, was to reason about the aircraft that were not available for inspection. Damage was recorded on returning planes across the fuselage and wings, with comparatively little around the engines. The inference is that hits to the engines were the ones that prevented a plane from returning, so armor belongs where the surviving planes show the least damage. It is worth saying plainly that the popular version, complete with a diagram of red dots, is a modern retelling; Wald's actual work estimated survival probabilities by hit location using a good deal more mathematics. The reasoning, though, is exactly right, and it generalizes.

The financial version is unglamorous and instructive. Databases of mutual fund returns routinely drop funds that close. Since funds mostly close after bad performance, the average return of the surviving list overstates what an investor would actually have earned. Analysts correct for this explicitly, and the correction is not small.

Now the entrepreneurship versions, which almost nobody corrects for:

• The dropout story. Several famous founders left college. To learn anything from that you would need the denominator: of everyone who left college to start a company, what fraction succeeded? That number is not in the book, because the book is about the people who succeeded.
• Advice books built from winners. A study picks companies that outperformed, looks for common traits, and publishes the traits as a recipe. The traits may be equally common among companies that failed, and nobody checked. Several celebrated business books have watched their exemplar companies subsequently stumble, which is what you would expect if the original selection captured luck along with skill.
• The halo effect. Once we know a company succeeded, we rate its culture as bold, its leadership as visionary, and its strategy as focused. When the same company falters, the same culture is described as arrogant, the leadership as autocratic, and the strategy as rigid. The attributes were inferred from the outcome, not measured independently.
• Era-locked advice. A founder who succeeded when customer acquisition on a particular platform was cheap will teach the tactic that worked. The tactic is not the lesson; the arbitrage is, and it closed.
• Your own network. The people you know who started businesses and are still around are, by construction, the survivors.

The correction is a single habit. Whenever someone tells you that successful founders do X, translate it into the only question that has information in it: of all the people who did X, what fraction succeeded, compared with those who did not do X? If nobody can answer that, you have been handed a story, not evidence. Stories are still useful, for vocabulary, for morale, for pattern recognition. They are just not evidence about your odds.

Key idea: Survivorship bias means reasoning from a sample filtered by success. The antidote is always to ask for the denominator: the fraction of everyone who tried that thing and succeeded.

Who actually becomes an entrepreneur

Popular imagination says entrepreneurs are young, exceptionally comfortable with risk, and born rather than made. The evidence is unkind to all three claims.

Risk tolerance. When researchers measure risk preference with standard instruments and compare entrepreneurs with managers of similar background, the differences are modest and inconsistent. What does show up reliably is something different: optimism about one's own case. In a large 1988 survey of nearly three thousand new business owners, Arnold Cooper, Carolyn Woo, and William Dunkelberg found that most respondents rated their own chances of success very high, with a third calling their own success certain, while rating the prospects of comparable businesses far lower. They were not embracing risk. They were, on average, not seeing it. That is a different psychological profile with a different remedy: not more caution in temperament, but more explicit base rates and more falsifiable tests, which is most of what this course teaches.

Age. The founder-as-twenty-two-year-old image collapses under data. Pierre Azoulay, Benjamin Jones, J. Daniel Kim, and Javier Miranda linked United States Census business records to founder ages and found that the average age of a person founding one of the very fastest-growing new firms was around forty-five. Success rates rise with founder age well into the fifties, and a fifty-year-old founder was roughly twice as likely as a thirty-year-old to build a top-performing firm. The mechanism is not mysterious. Industry experience, management experience, professional networks, and capital all accumulate with time, and they are exactly the inputs that predict survival.

Background and capital. Two of the strongest predictors of who starts a business are household wealth and having a self-employed parent. This is uncomfortable but important. The ability to go without income for a year, to sign a personal guarantee, or to absorb a failed venture is not evenly distributed, and advice to "just take the leap" quietly assumes a cushion that many people do not have. The practical response is not despair; it is to build the venture in a way that matches your actual cushion, which usually means starting part time, keeping fixed costs low, and reaching revenue early.

Necessity and opportunity. The Kauffman Foundation's indicators distinguish new entrepreneurs who came from a job or from choice from those who came from unemployment. The opportunity share rises in strong labor markets and falls in recessions, which is a reminder that a rise in business starts is not automatically good news about the economy.

Key idea: Entrepreneurs are not reliably more risk-tolerant, but they are reliably more optimistic about their own case; the average high-growth founder is around forty-five; and access to a financial cushion strongly shapes who can try at all.

Does entrepreneurship pay?

Here is the question nobody in the genre asks. Barton Hamilton examined United States self-employment earnings and found that the median self-employed person earns less than that same person would likely earn in paid employment, and the gap does not close with years of tenure. Related work by Tobias Moskowitz and Annette Vissing-Jorgensen found that returns to private business ownership as an asset class were not obviously better than public equity, despite far greater risk and no diversification. They called it the private equity premium puzzle.

Why do people do it anyway? Partly the nonpecuniary payoff, which is real: autonomy, meaning, building something, choosing your colleagues. Partly the shape of the distribution, since mean earnings are pulled up by a right tail even when the median is below wage work. And partly, per Cooper and colleagues, because most people think they are the right tail.

None of this is an argument against starting a business. It is an argument for starting one with your eyes open, for treating the decision as a real comparison against your best alternative, and for insisting on evidence before you commit money you cannot lose. That posture is what separates a founder from a gambler, and it is compatible with enormous ambition.

Key idea: On average, self-employment pays a median discount rather than a premium; the payoff is concentrated in a right tail plus nonpecuniary benefits, so compare honestly against your best alternative.

Try it

A widely shared post claims: "Eighty percent of founders who raised money had a technical cofounder, so get a technical cofounder." Identify the reasoning error and state what you would need to know to evaluate the advice.

Answer: The sample is filtered by success, since it includes only founders who raised money. The claim would only support the advice if you knew the fraction of all teams with a technical cofounder that raised money, compared with the fraction of all teams without one. Suppose ninety percent of applicants had a technical cofounder; then having one is associated with a lower success rate, and the same statistic would be evidence against the advice. Without the denominator the number carries no information about what you should do.

Common misconceptions

• "Ninety percent of businesses fail in the first year." BLS data show roughly 80 percent survive the first year; the ten-year figure, around a third surviving, is the sobering one.
• "A closure means a failure." Retirements, sales, and voluntary wind-downs all appear as closures, and many closed businesses were profitable.
• "Entrepreneurs are risk lovers." Measured risk tolerance differs little from comparable managers; measured optimism about one's own odds differs enormously.
• "Young founders have the advantage." Average age at the founding of the fastest-growing new firms is around forty-five, and success rates rise with age into the fifties.
• "Successful founders' habits are a recipe." Without knowing how common those habits were among failed founders, the list is a description of survivors, not a cause of survival.
• "Base rates do not apply to me." They apply as a starting point. You move your odds by industry choice, experience, capital cushion, and testing, not by exempting yourself from arithmetic.

Recap

• About 80 percent of new establishments reach one year, roughly half reach five, and about a third reach ten; the pattern is stable across decades.
• Closure data mixes failures with retirements, sales, and voluntary exits, so the survival curve overstates disaster and says nothing about owner satisfaction.
• Venture outcomes follow a power law: a large majority of investments return less than the capital invested and a few return more than ten times it.
• Survivorship bias is reasoning from a success-filtered sample; the antidote is to demand the denominator.
• Entrepreneurs are not notably more risk-tolerant but are notably more optimistic about their own case, and the average high-growth founder is near forty-five.
• Median self-employment earnings run below comparable wage earnings, which makes an honest comparison against your best alternative part of the work.

Sources

1. U.S. Bureau of Labor Statistics. (2025). Entrepreneurship and the U.S. economy: Business employment dynamics. bls.gov
2. U.S. Bureau of Labor Statistics. (2025). Business Employment Dynamics home page. bls.gov
3. Azoulay, P., Jones, B. F., Kim, J. D., & Miranda, J. (2018). Age and high-growth entrepreneurship (NBER Working Paper No. 24489). National Bureau of Economic Research. nber.org
4. Wikipedia contributors. (2025). Survivorship bias. en.wikipedia.org
5. Ewing Marion Kauffman Foundation. (2025). Kauffman indicators of entrepreneurship. kauffman.org

The big picture

The origin story of a business almost always gets rewritten. In the retelling there is a moment: a frustrating experience, a flash of insight, a napkin. In the actual record there is usually something duller and more useful. Somebody worked in an industry for six years, watched the same problem cost their employer money every quarter, noticed that the available solutions were bad, and eventually decided to go do it properly. That is not a story that sells a keynote. It is, however, how most durable businesses begin.

This matters because if you believe ideas arrive as lightning, your strategy for getting one is to wait. If you believe ideas are excavated from accumulated knowledge about a specific domain, your strategy is to go accumulate some, which you can start doing this week. The evidence favors the second view strongly enough that it should change what you do.

What the research says about origins

The most quoted evidence comes from Amar Bhide, who studied the founders of companies on the Inc. 500 list of fast-growing private firms. He found that a large majority, on the order of seven in ten, had replicated or modified an idea they encountered in previous employment. Only a small minority, a few percent, said their idea came from systematic search or formal research. Most had no business plan at the time of founding, and many had done no market research beyond talking to people they already knew.

Do not read that as an argument against planning; Bhide's founders were operating with an enormous informational advantage that substituted for research. They already knew the customers, the suppliers, the price points, and where the existing solutions failed. If you lack that advantage, research is how you buy some of it, which is what Module 2 is about.

Scott Shane's work generalizes the finding. People discover opportunities that sit close to knowledge they already have, from work, education, hobbies, or life circumstance. Give the same technology to three people and each will see a different application, shaped by what they already know about markets, customers, and how to serve them. This is why the advice to "find a problem you care about" is usually less useful than the advice to "find a problem you already understand better than almost anyone."

Key idea: Most founders get their idea from prior employment in the industry, not from search or epiphany. Domain knowledge is the raw material of opportunity recognition, and it is accumulable.

Drucker's sources of opportunity

Peter Drucker, writing in 1985, offered a checklist of places where opportunities systematically appear. It has aged well, in part because it points you at structural changes rather than at cleverness.

• The unexpected. An unexpected success, failure, or outside event. A product selling well to a customer you did not target is data, not noise. Ask why before you correct it.
• Incongruities. A gap between how something is and how it obviously should be. Anywhere a whole industry accepts a bad experience because that is how it has always been.
• Process need. A specific missing link that everyone in a workflow works around. These are the highest-conversion opportunities because the customer can already describe the problem for you.
• Industry and market structure change. Consolidation, deregulation, a new distribution channel, a dominant supplier losing power.
• Demographics. The most reliable source, because the data are published years in advance. An aging population guarantees demand for home care, mobility, and hearing services; you can read the projections today.
• Changes in perception. The facts stay the same and the meaning changes. Nothing about eggs changed when they moved from dangerous to healthy and back.
• New knowledge. The glamorous source, and Drucker's warning is that it has the longest lead times and the highest failure rates. Scientific breakthroughs take years to convert into a business, and the pioneer frequently loses to a later entrant with better distribution.

Notice what this list is not: it is not a list of technologies. Five of the seven sources are changes in the world around you that create demand whether or not anything was invented. That is where most workable small business opportunities live.

Key idea: Opportunities appear at points of change: unexpected results, incongruities, missing process links, structural shifts, demographics, changes in perception, and new knowledge, in roughly that order of practicality.

Problem-first and solution-first

Problem-first means starting from an observed, painful, recurring problem and searching for a solution. The advantage is enormous: demand already exists, and the customer can tell you what they currently spend to work around it. Most services businesses and most business-to-business software begin this way.

Solution-first means starting from a capability, often a technology or a skill, and searching for a problem it fits. This gets mocked as a solution in search of a problem, and the mockery is sometimes fair. But it is also how a great deal of real innovation happens, because some capabilities are genuinely new and nobody could have asked for them. The pharmaceutical industry is solution-first by construction. So was the laser, which spent years as an invention looking for applications before it landed in barcode scanners, surgery, and fiber optics.

The distinction that actually matters is not which end you start from. It is whether you have closed the loop. A solution-first founder who spends two months interviewing eleven kinds of potential user until she finds the one whose workflow her capability transforms has closed the loop honestly. A problem-first founder who identifies a real problem and then builds whatever he already knew how to build has not. Closing the loop means you can name a specific customer, the specific problem, and why your specific approach beats what they do today.

Key idea: Solution-first is not a sin; failing to search deliberately for the matching problem is. Either direction must end with a named customer, a named problem, and a defensible reason your approach wins.

From idea to opportunity

An idea is a sentence. An opportunity is a set of six answers. Write yours out and the weak link becomes obvious in about ten minutes.

Question	Weak answer	Strong answer
Who exactly?	Small businesses	Independent dental practices with two to five chairs and no in-house bookkeeper
What problem, how often?	Bookkeeping is annoying	Monthly reconciliation takes the office manager six hours and is late enough that the owner cannot see profitability until the quarter ends
What do they do today?	Nothing good	A local bookkeeper at 350 dollars a month, plus a spreadsheet the office manager maintains
Why you?	I am motivated	I ran billing for a four-location dental group for five years and know the insurance-reimbursement mess that makes generic bookkeeping fail here
How do you reach them?	Social media	State dental association meetings, the two practice-management software user groups, and referrals from dental CPAs
Does the arithmetic work?	It should be profitable	400 dollars a month, roughly 90 dollars of delivery cost per client, so about 310 dollars of contribution; I need 30 clients to replace my salary

Most raw ideas die at row five or row six, and that is the point. It is far cheaper to discover that you have no channel now than after you have spent a year building.

Key idea: An opportunity is a specific customer, a recurring problem with a current workaround, a reason it should be you, a reachable channel, and arithmetic where price exceeds cost by enough to matter.

Where to look, concretely

If you do not yet have a domain, here are the places opportunities keep turning up, ordered from most to least accessible.

• Your current job. Every workplace contains problems that cost money and that outsiders cannot see. Keep a running list for three months. Be careful about employment agreements and anything your employer owns; that is a real legal question and Module 5 covers where a lawyer becomes necessary.
• Geographic transplant. A model that works well in one metro and does not exist in yours. Unglamorous, frequently profitable, and the reason your town eventually gets a good bagel shop.
• Regulatory change. New rules create compliance work, reporting requirements, and demand for people who understand them. Read the trade press of any regulated industry and the opportunities announce themselves.
• Cost-curve shifts. When an input drops sharply in price, applications that were uneconomic become viable. Ask what you would build if this input were a tenth of today's price, then check how fast it is falling.
• Segments too small for incumbents. A category leader with a billion dollars of revenue cannot chase a fifteen-million-dollar niche. You can.
• The boring businesses. Commercial cleaning, HVAC, staffing, pest control, bookkeeping, landscaping, medical billing. Fragmented, demand-stable, cash-generating, and starved of owners who will answer the phone. These businesses do not trend, and they pay mortgages.

The secrecy question

New founders often spend their first weeks worrying that someone will steal the idea. Two honest halves to this. The first half: raw ideas are abundant and execution is scarce, most people are busy with their own lives, and investors will generally decline to sign a nondisclosure agreement because they see many similar pitches and cannot take on that liability. Refusing to describe your business will cost you far more in feedback and customers than it saves.

The second half, which the popular version of this advice omits: some things genuinely warrant protection. A formulation, a manufacturing process, a customer list, a piece of source code, or a pending patent application can be real trade secrets, and public disclosure can permanently destroy your ability to patent an invention abroad. The workable rule is to talk freely about the problem and the customer, and to be deliberate about disclosing the method. Module 5 covers what patents, trademarks, copyrights, and trade secrets actually protect, and where you need a licensed attorney rather than a rule of thumb.

Key idea: Talk openly about the problem you are solving; be deliberate about disclosing the method. Secrecy about the concept usually costs more feedback than it protects.

Try it

Take the raw idea: "an app that helps people find a therapist." Turn it into an opportunity statement using the six rows, or explain which row defeats it.

Answer: Rows one and three are the problem. "People" is not a customer; adults with commercial insurance seeking outpatient therapy in a specific metro is closer. What they do today is search their insurer's directory, call eight numbers, and find that most are not accepting patients, which is a real and describable pain. Row five is where most versions of this idea die: reaching patients one at a time is expensive, and the party with money and an existing channel is the insurer or the employer, not the patient. That observation usually converts the idea into a different and more viable business selling to employers or health plans. That conversion is the exercise doing its job.

Common misconceptions

• "Good ideas come from inspiration." They mostly come from working in an industry long enough to see its recurring problems clearly.
• "You need an original idea." Most fast-growing firms replicated or modified something the founder saw at a previous job.
• "Solution-first is always wrong." It is how pharmaceuticals, lasers, and much deep technology reached market. The requirement is a deliberate, honest search for the matching problem.
• "Passion is the main input." Passion sustains effort; domain knowledge generates the opportunity. Passion for a market you do not understand mostly generates expensive lessons.
• "Protect the idea above all." The concept is rarely the scarce asset, and secrecy about it costs feedback. Specific methods, formulas, and code are a different matter.
• "New technology is the best source of opportunity." Drucker ranked it last for practicality: longest lead times, highest failure rates, and pioneers often lose to later entrants with better distribution.

Recap

• Roughly seven in ten founders of fast-growing firms took their idea from prior employment; systematic search accounted for very few.
• Opportunity recognition depends on prior knowledge, which is why domain experience beats generic creativity.
• Drucker's sources are the unexpected, incongruities, process needs, structural change, demographics, changes in perception, and new knowledge.
• Problem-first and solution-first are both legitimate; what matters is closing the loop to a named customer, a named problem, and a defensible advantage.
• An opportunity statement answers six questions, and most raw ideas die on the channel question or the arithmetic.
• Discuss the problem openly and disclose methods deliberately; investors typically will not sign nondisclosure agreements.

Sources

1. Bhide, A. (1994). How entrepreneurs craft strategies that work. Harvard Business Review, 72(2), 150-161. hbr.org
2. Wikipedia contributors. (2025). Peter Drucker. en.wikipedia.org
3. Shepherd, D. A., et al. (2020). Identifying entrepreneurial opportunity. In Entrepreneurship. OpenStax, Rice University. openstax.org
4. U.S. Small Business Administration. (2025). Market research and competitive analysis. sba.gov
5. SCORE Association. (2025). Business planning and financial statements resources. score.org

The big picture

Here is the most common way a first venture dies, and it is almost gentle. You describe your idea to thirty people. Twenty-six say it sounds great. Several say they would definitely use it. Two offer to be your first customers. You quit your job, spend nine months and eighteen thousand dollars building it, launch, and email everyone. Four people click. Nobody pays. The two who offered to be first customers stop replying, and you cannot even be angry with them, because they were not lying. They were being kind about a hypothetical.

Customer discovery is the discipline of getting information out of those conversations instead of encouragement. The change is mostly mechanical: different questions, different interpretation, different record-keeping. It is learnable in an afternoon and it is the single highest-return skill in this course, because everything downstream, your product, your pricing, your channel, is built on what you believe about the customer.

Why enthusiasm is worthless

Three forces conspire to make people say nice things about your idea.

The first is ordinary politeness. You are visibly invested. Telling you your idea is weak imposes a social cost on the speaker for no benefit, so almost nobody does it. Researchers call the general pattern social desirability bias, and it is strongest exactly where you most need honesty.

The second is that hypothetical future behavior is free to promise. "Would you use this?" costs the respondent nothing to answer yes. Stated intentions and revealed behavior diverge routinely, and the gap widens for anything involving future effort, future money, or a change of habit. Anyone who has run a gym in January has seen the data.

The third is that you are asking leading questions without noticing. "Do you find scheduling frustrating?" contains its own answer. Almost everyone finds almost everything mildly frustrating when asked. The information you needed was whether they find it frustrating enough to have already done something about it, and that question sounds completely different.

Steve Blank's formulation of customer development, developed after his own experience running startups that built products nobody wanted, reduces to a slogan worth taking literally: the facts are not inside your building. No amount of internal debate produces information about customers. Only customers do, and only if you ask them properly.

Key idea: Politeness, free hypotheticals, and leading questions make enthusiasm the default response to any idea. Positive reactions are therefore not evidence.

The question rewrite

Rob Fitzpatrick's rule of thumb is that a good question is one you could ask your own mother and still get useful data from, because the answer is a fact about her life rather than an opinion about your idea. Three habits follow: talk about their life, not your idea; ask about specifics in the past, not generalities about the future; talk much less than you think you should.

Weak question	Why it fails	Rewrite
Would you use an app that schedules your crews automatically?	Hypothetical, leading, free to answer yes	Walk me through how you built last week's schedule. What did you use, and how long did it take?
Would you pay 200 dollars a month for this?	Invites a polite number with no consequence	What do you spend on scheduling today, counting software and the hours it takes? Who approves that spending?
Do you think this is a good idea?	Asks for an opinion about you, not a fact about them	What have you already tried to fix this, and why did you stop?
How often do you have scheduling problems?	Generic, invites an estimate rather than a memory	When did it last go wrong? What happened? And the time before that?
Don't you hate how the current tools work?	Loaded with the answer	Tell me about the tool you use now. What is the last thing that annoyed you about it?

Notice that every rewrite asks for a story from the past. Past behavior is a fact; the respondent either did the thing or did not, and the details come with a texture that made-up answers rarely have. When someone says the reconciliation took four hours last Tuesday and describes which spreadsheet tab broke, you have learned something. When someone says they would probably use a tool like that, you have learned that they are polite.

Key idea: Replace every hypothetical and leading question with a request for a specific story from the recent past, then follow it with what they did about it and what it cost.

The commitment ladder

Signals differ enormously in what they cost the person giving them, and cost is what makes a signal informative. Rank everything you hear on this ladder.

Signal	Cost to them	Evidence value
"That is a great idea"	None	None
"Send me more information"	Almost none	Very weak
Joins a waiting list with an email address	Trivial	Weak
Introduces you to a colleague by name	Reputation	Moderate, they are staking credibility
Gives you an hour with their operations team, or shares real data	Time and access	Moderate to strong
Signs a nonbinding letter of intent	Small but real	Strong for enterprise sales
Prepays, or puts down a deposit	Money	Very strong
Pays, uses it repeatedly, and renews	Money plus habit change	Definitive

A useful discipline is to end interviews with an ask that is one rung higher than you are comfortable with. Not "does this sound interesting" but "if I have a working version in six weeks, would you run a paid pilot at four hundred dollars a month?" You will get fewer yeses. Every one of them will be worth something.

Compliments are the fool's gold here. A conversation that produced three compliments and no commitments and no facts produced nothing. Write that in your notes explicitly, because the memory of a warm conversation will otherwise get filed as evidence.

Key idea: Weigh every signal by what it cost the person to give. Compliments cost nothing; introductions cost reputation; deposits cost money; renewal costs habit.

Running the program

How many. Three interviews tell you nothing; you will hear three idiosyncratic stories. Five hundred is procrastination dressed as rigor. In practice, within a single well-defined segment, patterns start repeating somewhere between fifteen and thirty conversations. The stopping rule is saturation: when you can predict the next person's answers before they give them, that segment is understood and you should either move to a new segment or start testing solutions. If you cannot predict answers after thirty interviews, your segment is probably too broad and needs splitting.

Who. Friends and family are the worst possible sample, because they are optimizing for your feelings. Recruit through trade associations, professional groups, online communities where your customers actually complain, conference hallways, cold outreach that asks for advice rather than a sale, and referrals from each interview. Ask every interviewee for two introductions; this snowball method works well but produces correlated samples, so deliberately seed several independent chains rather than one.

Structure. Thirty minutes. Two minutes of framing in which you say plainly that you are researching a problem and are not selling anything, which is both honest and disarming. Twenty minutes of their story, following the rewritten questions. Five minutes asking them to rank the problems they have mentioned and say which one they would pay to remove. Three minutes for referrals and permission to follow up. If you want to show your concept, do it after all of that, label it as a pitch, and treat the reaction as marketing research rather than problem research.

Notes. Write verbatim quotes wherever you can. Separate facts (what they did, what they spend, what tool they use) from opinions (what they think of your idea). Log three fields for every interview: the current workaround, the money and hours it consumes, and the highest rung of the commitment ladder reached. Twenty interviews with those three fields filled in will tell you more than a hundred pages of impressions.

Key idea: Interview fifteen to thirty people in one segment, recruit outside your social circle, follow a fixed structure, and record current workaround, current cost, and commitment level for each.

What discovery cannot do

Two honest limits. First, discovery can kill a bad idea quickly and can sharpen your understanding of a segment enormously, but it cannot prove demand. Nothing short of money proves demand, which is why Module 3 moves from interviews to experiments and paid pilots. Treat discovery as the cheap filter that stops you from building the obviously wrong thing.

Second, customers are experts on their problems and amateurs at your solution. They can tell you precisely what goes wrong on Tuesday and what they currently pay to survive it. They generally cannot design the product, and asking them to is a way of outsourcing a judgment that is yours. There is a famous quotation attributed to Henry Ford about customers asking for faster horses, and it is worth knowing that no reliable source for it has ever been found. The underlying point survives the missing citation, but so does its abuse: founders quote it to justify ignoring customers entirely, which is the opposite of the lesson. Listen carefully to the problem. Decide the solution yourself. Then test it.

One more asymmetry worth planning around. Business customers are usually easier to interview well, because they can name a budget, a decision-maker, and a cost of the problem in dollars. Consumers are harder, because people misreport their own habits, especially around money, health, and time. If you are building for consumers, weight behavioral experiments more heavily and interviews less.

Key idea: Discovery filters out bad ideas and sharpens segments; only payment demonstrates demand. Customers are authoritative about problems and unreliable about solutions.

Try it

Rewrite these three questions, then say what fact each rewrite is trying to obtain. (1) "Would you buy organic dog treats if they were locally made?" (2) "Do you think my pricing is fair?" (3) "How much would you pay for a service that files your quarterly taxes?"

Answer: (1) "What treats did you buy last time, where, and how did you choose?" The fact sought is actual purchase behavior and the decision criteria used, including whether local sourcing ever entered the decision. (2) "What are you paying for the closest alternative today, and what would have to be true for you to switch?" The fact sought is the current reference price and the switching threshold, since fairness is an opinion and the reference price is a fact. (3) "Walk me through what happened at the last quarterly filing. Who did it, how long did it take, and what did it cost?" The fact sought is the current spend in money and hours, which is the only defensible anchor for pricing.

Common misconceptions

• "Positive feedback validates the idea." Politeness produces positive feedback for nearly every idea; only costly signals carry information.
• "Asking would you buy this is market research." It measures willingness to be agreeable, not willingness to pay.
• "More interviews are always better." Beyond saturation within a segment, additional interviews mostly delay the experiments that would actually test demand.
• "Friends give useful early feedback." They are the most biased sample available and the least likely to tell you the idea is weak.
• "Customers will tell you what to build." They are authoritative on their problems and unreliable on solutions; the design judgment stays yours.
• "Ford proved that customers do not know what they want." The faster-horses quotation has no verified source, and it is routinely used to excuse not listening at all.

Recap

• Politeness, free hypotheticals, and leading questions make enthusiasm meaningless as evidence.
• Good questions ask for specific stories from the recent past, including what the person did and what it cost.
• The commitment ladder ranks signals by their cost to the giver, from compliments through introductions and deposits to renewal.
• Fifteen to thirty interviews within one clearly defined segment usually reach saturation; recruit outside your social circle and seed several referral chains.
• Record the current workaround, the money and hours it consumes, and the highest commitment reached, for every interview.
• Discovery kills bad ideas cheaply but cannot prove demand, and customers are experts on problems rather than on your solution.

Sources

1. Blank, S. (2013). Why the lean start-up changes everything. Harvard Business Review, 91(5), 63-72. hbr.org
2. Wikipedia contributors. (2025). Customer development. en.wikipedia.org
3. U.S. Small Business Administration. (2025). Market research and competitive analysis. sba.gov
4. Shepherd, D. A., et al. (2020). Customer discovery and problem solving. In Entrepreneurship. OpenStax, Rice University. openstax.org
5. SCORE Association. (2025). Conducting customer research. score.org

The big picture

There is a slide that appears in an enormous share of first business plans. It says the market is worth fifty billion dollars a year, and that capturing just one percent of it produces five hundred million dollars of revenue. Everyone who reads business plans for a living has learned to stop reading at that slide, and it is worth understanding exactly why, because the reason is not snobbery.

The problem is that the slide contains no information. It does not say who buys, how they are reached, how many you can serve, or what would have to be true for the one percent to happen. It works backwards from a desirable answer to an assumption that produces it. Meanwhile, the honest version of the same analysis, which counts customers upward from things you can verify, usually reveals that the market is smaller than the report said, that you cannot serve most of it, and that your real constraint is something else entirely. That last discovery is the valuable one.

Three nested numbers

The standard vocabulary has three layers, and being sloppy about which one you are quoting is how founders end up misleading themselves and their investors at the same time.

• TAM, the total addressable market. Everyone in the world who could conceivably buy something in this category, if there were no constraints on geography, channel, or product fit. Total annual revenue if you owned the entire category.
• SAM, the serviceable available market. The slice your actual product, in your actual geography, sold through your actual channel, could serve. Excluding the parts you cannot reach is not pessimism; it is the difference between a plan and a wish.
• SOM, the serviceable obtainable market. What you can realistically win in a defined period, given your capacity, your sales effort, and the competitors already in the room. This is the only one of the three that belongs in a revenue forecast.

There are two ways to compute these. Top-down starts with a published industry number and shaves it with percentages. It is fast and almost always wrong, because published market definitions rarely match your product and because every shaving percentage is invented. Bottom-up counts units: how many customers exist, how often they buy, at what price. It is slower, it forces you to name assumptions, and it produces a number you can defend line by line. Build bottom-up. Use top-down only as a sanity check on the order of magnitude.

Key idea: TAM is the category, SAM is what your product and channel can serve, and SOM is what you can win. Build all three bottom-up from countable units, and put only SOM in the forecast.

A bottom-up sizing, worked

Say you are considering a mobile dog-grooming business: a fitted van that comes to the driveway, in a metro area of about 1.2 million people. Here is the build, with every assumption labeled as either a published figure or a guess.

Step	Calculation	Result	Source of assumption
Households	1,200,000 people / 2.5 per household	480,000	Census average household size
Dog-owning households	480,000 x 45 percent	216,000	AVMA pet ownership survey
Dogs needing professional grooming	216,000 x 40 percent	86,400	Guess, from coat types; needs checking
Willing to pay a premium for at-home service	86,400 x 15 percent	12,960 households (SAM)	Guess; the weakest number here
Annual spend per household	6.5 grooms per year x 110 dollars	715 dollars	Local salon pricing plus a mobile premium
SAM in revenue	12,960 x 715	About 9.3 million dollars	Derived

Now the step most people skip, which is where the plan actually gets made. What can one van do? A groomer can handle about five dogs a day, five days a week, forty-eight weeks a year: 1,200 grooms annually. At 6.5 grooms per client per year, that is about 185 recurring clients, and 1,200 times 110 dollars is 132,000 dollars of annual revenue per van.

So a single van captures 185 of 12,960 households, or about 1.4 percent of the SAM. Three vans by year three would serve roughly 555 households, about 4.3 percent of the SAM, producing around 396,000 dollars of revenue.

Here is the punchline. Run the sensitivity: suppose only 8 percent of grooming-needing households want mobile service instead of 15 percent. The SAM collapses from 9.3 million to about 4.9 million dollars. Does the three-van plan change? Not at all. You still cannot groom more than 1,200 dogs per van per year, and 555 households out of 6,912 is still under 9 percent of the market. Market size was never the binding constraint. Capacity was, and behind capacity sits the real question: can you hire, train, and retain three competent groomers in this metro, and can you keep 555 households on a schedule?

That reframing is what a bottom-up sizing is for. To reach two million dollars of revenue you would need roughly fifteen vans and fifteen groomers, and the honest plan is now a discussion about recruiting in a trade with chronic labor shortage, not about a nine million dollar market.

Key idea: Size the market bottom-up, then immediately test your capacity to serve it. If halving the market assumption does not change your plan, market size is not your constraint and you should go find the one that is.

When market size genuinely matters

None of this means TAM is useless. It matters in three specific situations. First, when raising venture capital, because a fund needs the possibility of a very large outcome to justify the risk, and a business capped at twenty million dollars of revenue cannot produce one no matter how well run. Second, when you must commit large fixed costs before revenue, such as writing software, tooling a factory, or pursuing a regulatory clearance, because those costs have to be spread over a big enough base. Third, when choosing between two markets to enter with the same capability, where relative size is a legitimate tiebreaker.

For free, credible top-down anchors, the government is better than most paid reports: the Census Bureau's County Business Patterns and Economic Census give establishment and revenue counts by industry and county, the Bureau of Labor Statistics gives employment and wages by occupation and area, and trade associations publish member counts. Anchoring on those and building upward beats quoting a headline from a press release about a report you have not read.

Competition, including the competitor you forgot

Founders often say, sometimes proudly, that they have no competitors. There are only three explanations, and none is good. Either there is no market, and people have declined to solve this problem because they do not care about it; or you have not looked hard enough; or you have defined the category so narrowly that the statement is true and meaningless, in the way that a restaurant is the only Peruvian-Korean fusion place on its block.

Build the competitor set in four layers:

• Direct competitors. Companies selling roughly what you sell to roughly your customer.
• Indirect substitutes. Different products that solve the same job. A grooming van competes with the salon, the mobile groomer's rival, and the bathtub.
• In-house alternatives. The customer doing it themselves, usually with a spreadsheet, a part-time employee, or an intern. In business software this is the most common competitor by far.
• Doing nothing. The status quo, which requires no purchase order, no training, no risk, and no meeting. In almost every category, doing nothing has the largest market share, and it never appears on the competitive matrix.

Michael Porter's five forces, rivalry, buyer power, supplier power, threat of new entrants, and substitutes, were designed to explain the average profitability of an industry, not the prospects of one new venture. Used for its intended purpose it is still valuable to a founder: it answers why this industry earns what it earns, and therefore what you would have to do differently to beat that average. If buyers are concentrated and switching costs are low, you should expect price pressure regardless of how good your product is, and your plan needs an answer for that.

Key idea: Your competitor set includes direct rivals, substitutes, in-house workarounds, and doing nothing, and the status quo is usually the market leader.

Being second is fine

The phrase first-mover advantage is repeated so often that it sounds like a law. The evidence is much weaker than the phrase implies. Research by Peter Golder and Gerard Tellis, re-examining market pioneers across dozens of categories, found that a large share of true pioneers failed outright and that market leadership frequently ended up with early followers rather than inventors. The pattern is easy to see in familiar cases: there were search engines before Google, social networks before Facebook, portable music players before the iPod, and electric cars long before any modern manufacturer. Later entrants won by solving the execution and distribution problems the pioneers had not.

Where first movers really do gain, it is through specific mechanisms rather than glory: network effects that make a product more valuable as it gains users, high switching costs, control of a scarce input, or a genuine learning curve. If your market has none of those, being early buys you nothing but the expense of educating customers for whoever comes third.

What you do need is differentiation that is legible to one specific segment. "Somewhat better for everyone" is the weakest possible position, because nobody switches for slightly. Pick a segment and be dramatically better for them on a dimension they can feel: faster, cheaper by a lot, specialized to their workflow, or available where nothing is. The grooming van is not competing on quality of haircut. It is competing on the fact that a person with mobility limits, three children, or a fearful dog does not have to leave the house.

Key idea: Pioneer advantage is weak in general and real only through network effects, switching costs, scarce inputs, or learning curves. Differentiation must be dramatic for a specific segment rather than mild for everyone.

Try it

Size the market, bottom-up, for a bookkeeping service aimed at independent restaurants in a state with roughly 9,000 restaurants, of which about 60 percent are independent rather than chain-operated, and of which perhaps a third would outsource bookkeeping at 500 dollars a month. Then find the binding constraint if one bookkeeper can carry 25 clients.

Answer: Independents: 9,000 times 0.60 equals 5,400. Willing to outsource: 5,400 times one third equals 1,800 restaurants. Annual spend: 500 dollars times 12 equals 6,000 dollars. SAM equals 1,800 times 6,000, which is 10.8 million dollars. Capacity: one bookkeeper serves 25 clients, generating 150,000 dollars of revenue. Reaching one million dollars of revenue means about 167 clients and seven bookkeepers, which is 9 percent of the SAM. The binding constraint is hiring and training seven qualified bookkeepers who understand restaurant accounting, plus the sales effort to add roughly four clients a month for three and a half years. The market size did not determine any of that.

Common misconceptions

• "A big TAM means a good business." Large markets attract large competitors, and your revenue is bounded by capacity and sales, not by the size of the category.
• "One percent of a huge market is a modest goal." One percent of a national market usually implies thousands of customers acquired through a channel you have not built.
• "Top-down sizing from an industry report is rigorous." Report definitions rarely match your product, and every shaving percentage in a top-down build is an unlabeled guess.
• "No competitors is a selling point." It signals no market, insufficient research, or an artificially narrow category definition.
• "Doing nothing is not a competitor." The status quo requires no purchase decision and usually holds the largest share of any category.
• "First movers win." Pioneers fail at high rates; advantage requires network effects, switching costs, scarce inputs, or learning curves.

Recap

• TAM is the whole category, SAM is what your product and channel can serve, and SOM is what you can realistically win.
• Bottom-up sizing counts customers, frequency, and price, labeling every assumption as published or guessed.
• In the grooming example, one van serves about 185 households for 132,000 dollars a year, so capacity rather than market size sets the plan.
• If halving a market assumption leaves the plan unchanged, market size is not the binding constraint.
• The competitor set is direct rivals, substitutes, in-house workarounds, and doing nothing, and Porter's forces explain industry-average profitability.
• First-mover advantage is unreliable; differentiation must be dramatic for a defined segment.

Sources

1. U.S. Census Bureau. (2025). County Business Patterns. census.gov
2. American Veterinary Medical Association. (2024). Reports and statistics on pet ownership. avma.org
3. U.S. Small Business Administration. (2025). Market research and competitive analysis. sba.gov
4. Wikipedia contributors. (2025). Total addressable market. en.wikipedia.org
5. Wikipedia contributors. (2025). First-mover advantage. en.wikipedia.org

The big picture

There is a gap between having a product and having a business, and a surprising number of ventures fall into it. The product works. People like it. And yet nothing adds up, because nobody ever wrote down how money would move: who pays, how they are reached, what it costs to serve them, and who else has to be involved for any of it to happen. A business model is the answer to that set of questions. It describes how a venture creates value, delivers it, and captures some of it as revenue.

This lesson gives you two tools and an honest account of their limits. The first is the value proposition, which forces you to state what you are actually selling in the customer's terms rather than yours. The second is the business model canvas, which has become the default whiteboard exercise in every entrepreneurship program in the world, is genuinely useful, and is also the most over-trusted artifact in the field. We will use it, and then we will say plainly what it cannot do.

The value proposition, and the job the customer is hiring you for

Most first attempts at a value proposition are a list of features written from inside the company. "Cloud-based, AI-powered, mobile-first inventory management with real-time dashboards." A customer reading that learns nothing about whether their Tuesday gets better.

The reframing that helps most is the jobs-to-be-done idea associated with Clayton Christensen: customers do not buy products, they hire them to make progress on a job in a particular circumstance. The famous illustration involves a fast food chain trying to sell more milkshakes. Segmenting by customer demographics and asking people what they wanted produced nothing. Watching who actually bought revealed that a large share sold before nine in the morning, to solo commuters, who were hiring the milkshake to make a long boring drive tolerable and to hold off hunger until lunch. In that circumstance the competition was not other milkshakes; it was bagels, bananas, and boredom, and the winning improvements were about thickness and how long the drink lasted rather than flavor variety.

Two honest notes about that story. It is a consulting case rather than a controlled study, and you should treat it as a memorable framing device, not as research evidence. And the framework has a real failure mode: any observed behavior can be narrated as a job after the fact, which makes it hard to falsify. Use it as a prompt for better questions, then verify with the interview methods from the last lesson.

A workable value proposition names four things: the customer, the job, the improvement, and the comparison. Compare these two:

• Weak: "We offer modern, streamlined bookkeeping for small businesses using best-in-class technology."
• Strong: "For independent dental practices with two to five chairs, we close the books by the fifth of each month instead of the twentieth, so the owner sees profitability while it can still be acted on, for 400 dollars a month against the 350 dollars they pay a bookkeeper plus six hours of the office manager's time."

The second version is checkable. A customer can disagree with it, which is exactly what makes it useful. Note that it quantifies the improvement and names the current alternative, because value is always relative to what the customer does today, never absolute.

Key idea: A value proposition states who the customer is, what job they are hiring you for, how much better you are than their current alternative, and in what units. If it cannot be disagreed with, it is marketing copy rather than a proposition.

The nine blocks

Alexander Osterwalder and Yves Pigneur's business model canvas organizes a business into nine blocks on one page. Here it is, filled in for the mobile dog-grooming van from the last lesson, because an abstract canvas teaches nothing.

Block	Question	Grooming van
Customer segments	Who are we serving?	Households with grooming-breed dogs; priority on owners with mobility limits, small children, or anxious dogs
Value propositions	What job, how much better?	A full groom without leaving home, no crate time, same groomer every visit, about 25 dollars above salon price
Channels	How do they find and buy?	Veterinarian referral cards, neighborhood social groups, van signage, online booking
Customer relationships	How is the relationship maintained?	Standing eight-week appointments booked on the spot, text reminders, one named groomer per route
Revenue streams	How does money arrive?	110 dollars per groom, prepaid packages of six at a small discount, add-on nail and teeth services
Key resources	What must we own or have?	Fitted van, water and power system, licensed groomers, booking software, insurance
Key activities	What must we do well?	Route density, scheduling, groomer recruitment and retention, handling difficult dogs safely
Key partnerships	Who do we depend on?	Veterinary clinics for referrals, van outfitter, supply distributor, mobile-business insurer
Cost structure	What does it cost?	Groomer wages, van payment and fuel, supplies, insurance, software, marketing

Filling this in takes twenty minutes and does something valuable: it exposes the blocks where you have nothing. Most founders discover they cannot answer channels, and channels are usually where the business lives or dies.

Key idea: The canvas is a completeness check. Its main service is showing you which of the nine questions you have never answered, and channels is the block most often empty.

What the canvas cannot do

Now the criticism, because the canvas is routinely treated as if a completed one were a validated business.

• It has no competition block. Nine blocks and not one asks who else is in the market or what the customer does today. You must import that from the previous lesson.
• It has no time dimension. A canvas describes a steady state that does not exist yet, and says nothing about sequence: which customers first, which channel first, what must be true before hiring.
• It does not evaluate. There is no test in the tool for whether the model is good. A canvas for a business that loses money on every sale looks identical to one for a business that mints money, until you do the arithmetic yourself, which is the next lesson.
• Sticky notes are cheap. Every block invites a plausible sentence, and plausible sentences feel like progress. A completed canvas is a set of guesses arranged neatly, and the neatness is the danger.
• It does not rank risk. All nine blocks look equally important on the page, but in any real venture one or two assumptions carry nearly all the risk.

Ash Maurya's Lean Canvas is a partial answer to some of this: it replaces four blocks with problem, solution, key metrics, and unfair advantage, which pushes you toward the riskiest parts. Either version works if you use it correctly, and the correct use has three steps. Fill in all nine blocks quickly. Then circle the two that would destroy the business if you are wrong about them. Then stop drawing and go test those two. For the grooming van, the two circled blocks are almost certainly key resources, since the entire business depends on hiring and keeping licensed groomers in a labor-short trade, and channels, since customer acquisition cost decides whether route density is achievable. Everything else can be adjusted later.

Key idea: Treat every block as a hypothesis, rank them by how badly being wrong would hurt, and test the top two. The canvas organizes your uncertainty; it does not reduce it.

Revenue models and their real arithmetic

The revenue streams block deserves its own treatment, because founders frequently choose a model by fashion rather than by fit. Here are the main options with the numbers that decide whether they can work.

Model	How it works	The arithmetic that decides it
One-time sale	Customer buys a product outright	Margin per unit times units; you must find new customers continuously
Subscription	Recurring fee for continuing access	You are now in the churn business: at 5 percent monthly churn, average customer life is 20 months
Usage-based	Pay per unit consumed	Aligns with customer value but makes revenue lumpy and forecasting hard
Freemium	Free tier converts a fraction to paid	Typical free-to-paid conversion is roughly 2 to 5 percent, and free users still cost money to serve
Marketplace take rate	Percentage of transactions between others	At a 15 percent take rate, one million dollars of revenue requires about 6.7 million dollars of transaction volume
Advertising	Sell attention to third parties	One million monthly pageviews with two slots each at a 5 dollar CPM is about 10,000 dollars a month
Services and retainers	Sell time and expertise	Billable hours times rate times utilization; scales linearly with headcount
Licensing and franchising	Others use your brand or technology	Requires a proven, documentable system before anyone will pay for it

Work the freemium line, because it is the one most often adopted without arithmetic. Suppose you convert 3 percent of free users. To reach 1,000 paying customers you need roughly 33,000 free users. If serving a free user costs even 20 cents a month in infrastructure and support, those 32,000 non-paying users cost about 6,400 dollars a month, which your 1,000 paying customers must cover before you earn anything. Freemium is a marketing expense disguised as a product tier, and it works when the free tier drives referrals or network effects and fails when it is simply a discount to people who were never going to pay.

Advertising deserves the same treatment. A million monthly pageviews sounds like success. At two ad slots per page and a five dollar cost per thousand impressions, it produces ten thousand dollars a month, before the cost of producing whatever attracts a million pageviews. Advertising is a model for very large audiences or very valuable niches, and almost never a first revenue model for a new venture.

Key idea: Choose a revenue model by arithmetic, not fashion. Freemium needs a conversion rate and a cost to serve free users; advertising needs enormous traffic; marketplaces need transaction volume many times their revenue.

Try it

A founder plans a subscription app at 8 dollars a month for home cooks, with a free tier. She projects 200,000 free users in year two and a 5 percent conversion. Check the plan.

Answer: 200,000 times 5 percent is 10,000 paying users, which at 8 dollars a month is 80,000 dollars of monthly revenue, or 960,000 dollars a year. That part is fine. The problems are elsewhere. First, 5 percent is at the top of the observed freemium range for consumer apps, and 2 percent is a more defensible planning figure, which would halve revenue to about 384,000 dollars. Second, the plan says nothing about acquiring 200,000 free users; at even one dollar of paid acquisition cost per free user that is 200,000 dollars of spend, and consumer app acquisition is often much more expensive. Third, consumer subscription churn is typically high, so the 10,000 paying users are not a stock but a leaky bucket. The canvas would have shown a filled revenue stream block and an empty channel block, which is precisely the failure pattern this lesson warns about.

Common misconceptions

• "A good product is a business model." A model also specifies who pays, how they are reached, what serving them costs, and who else must be involved.
• "A completed canvas validates the idea." It arranges guesses neatly. Validation comes from testing the riskiest blocks.
• "The canvas covers everything important." It has no block for competition, no time dimension, and no evaluation of whether the model earns money.
• "Jobs-to-be-done is settled research." The milkshake case is an illustrative consulting story, and the framework is easy to apply unfalsifiably after the fact.
• "Freemium is free marketing." Free users consume support and infrastructure, and conversion is typically a few percent.
• "Advertising will cover costs once we have traffic." A million monthly pageviews at typical rates produces roughly ten thousand dollars a month.

Recap

• A business model specifies how value is created, delivered, and captured, not just what the product does.
• A strong value proposition names the customer, the job, a quantified improvement, and the current alternative it beats.
• The canvas's nine blocks are customer segments, value propositions, channels, customer relationships, revenue streams, key resources, key activities, key partnerships, and cost structure.
• Its blind spots are competition, time, evaluation, and risk ranking; the fix is to circle the two riskiest blocks and test them.
• Revenue models must be chosen by arithmetic: freemium needs conversion and cost-to-serve, marketplaces need volume, advertising needs scale.
• The most commonly empty block is channels, and channels usually decide whether the business works.

Sources

1. Wikipedia contributors. (2025). Business Model Canvas. en.wikipedia.org
2. Christensen, C. M., Hall, T., Dillon, K., & Duncan, D. S. (2016). Know your customers' jobs to be done. Harvard Business Review, 94(9), 54-62. hbr.org
3. Shepherd, D. A., et al. (2020). Business model canvas. In Entrepreneurship. OpenStax, Rice University. openstax.org
4. U.S. Small Business Administration. (2025). Write your business plan. sba.gov
5. Strategyzer. (2025). Business model canvas resources. strategyzer.com

The big picture

Walk into a bakery at nine on a Saturday morning with a line to the door and you will assume the owner is doing well. Quite often she is not. Busy and broke is one of the most common conditions in small business, and it is always the same arithmetic: the price is too low, the variable cost is higher than she thinks, or the fixed costs quietly grew until no achievable volume covers them. None of that is visible from the line at the counter. All of it is visible in about twenty minutes of calculation.

This lesson is that calculation. It is the most numerical part of the course and the most directly useful. If you learn nothing else from this course, learn to compute a contribution margin, a break-even, and a payback period, because those three numbers will tell you more about a business than any amount of enthusiasm.

Contribution margin

Variable costs change with each unit you sell. Fixed costs do not, at least in the short run. Contribution margin is price minus variable cost per unit, and it is the amount each sale contributes toward covering fixed costs and, eventually, profit.

Take a bakery selling a loaf of sourdough at 8 dollars.

Item	Amount	Note
Price	8.00	Retail, over the counter
Ingredients	1.80	Flour, salt, water, starter, energy for the oven
Packaging	0.25	Bag and label
Card processing	0.53	2.9 percent of 8.00 plus 30 cents, assuming card payment
Variable cost	2.58
Contribution margin	5.42	67.75 percent of price

Notice the card fee. Founders leave payment processing out of variable cost constantly, and on a low-priced item the fixed 30 cent component is brutal: on a 3 dollar coffee it is nearly ten percent of the price before the percentage fee. Now the fixed costs, per month:

Rent 3,800; utilities 900; insurance 250; equipment loan payment 700; one employee at 3,200 fully loaded including payroll taxes; software, accounting, and miscellaneous 350. Total fixed cost is 9,200 dollars a month.

Break-even units equals fixed costs divided by contribution margin: 9,200 divided by 5.42 is about 1,698 loaves a month. Spread over 26 operating days, that is 65 loaves a day before the business earns a cent.

Now add the number the owner left out, which is her own pay. Suppose she needs 4,000 dollars a month to live. Fixed costs become 13,200, break-even becomes 13,200 divided by 5.42, which is about 2,436 loaves a month, or 94 loaves a day. Adding the owner's modest salary raised the daily requirement by 45 percent. This is the single most common error in small business planning: treating the founder's labor as free, which makes a business that cannot support its owner look like a business that breaks even.

You can also express break-even in revenue: fixed costs divided by contribution margin percentage, or 13,200 divided by 0.6775, which is about 19,483 dollars of monthly sales. Whichever form you use, the last step is the reality check. Can a 400 square foot shop with one employee actually sell 94 loaves a day, six days a week, in this neighborhood? If the honest answer is no, you have learned something enormously valuable for the price of an afternoon.

Key idea: Break-even units equal fixed costs divided by contribution margin. Always include the owner's pay in fixed costs; leaving it out is how unviable businesses look viable.

Pricing, and the arithmetic of changing it

There are three ways to set a price and only one of them is usually right.

• Cost-plus. Add a markup to your cost. Simple, common, and weak: it prices your own cost structure rather than the customer's alternative, it rewards inefficiency with a higher price, and it leaves value on the table whenever what you provide is worth much more than it costs you.
• Competitive. Price near what rivals charge. A reasonable anchor and a poor strategy, because it silently assumes you are the same as your competitors, which if true is a bigger problem than pricing.
• Value-based. Price against the customer's next best alternative and the value you create relative to it. If your service saves a dental practice six office-manager hours a month and replaces a 350 dollar bookkeeper, the reference value is roughly 350 dollars plus six hours of wages, and your price lives in relation to that number, not to your own costs.

New founders systematically underprice. Some of it is fear of rejection, some is an accurate sense that their offering is unproven, and some is the belief that low price is a strategy. It rarely is, for two reasons. Price is a quality signal in categories where quality is hard to judge in advance, and cheap customers are frequently the most expensive to serve. Raising prices often improves the business twice: better margin per customer, and the departure of the customers who consumed the most support.

Before changing a price, do the volume arithmetic. Suppose a product sells at 100 dollars with 60 dollars of variable cost, so contribution margin is 40 dollars, or 40 percent.

The discount. Cut the price 20 percent to 80 dollars. Variable cost is unchanged, so contribution margin falls to 20 dollars. Each sale now contributes half as much, which means you need to sell twice as many units to end up in the same place. A 20 percent discount on a 40 percent margin demands a 100 percent volume increase. Very few discounts achieve that, which is why casual discounting destroys more small businesses than competition does.

The increase. Raise the price 10 percent to 110 dollars. Contribution margin rises to 50 dollars, a 25 percent improvement per unit. How many customers can you lose and still be even? You break even at a 20 percent unit loss, since 0.80 times 50 equals 40. In most markets, a 10 percent price increase does not cost you one customer in five. This asymmetry is why pricing is the highest-leverage variable in a small business, and why so few owners touch it.

Key idea: Required volume increase from a discount equals the discount divided by the difference between margin percentage and discount percentage. At a 40 percent margin, a 20 percent discount needs double the volume, while a 10 percent increase survives a 20 percent customer loss.

CAC, LTV, and payback

Customer acquisition cost is what you spend to get one new customer. The only defensible version is fully loaded: all sales and marketing spending in a period, including salaries, commissions, tools, agency fees, and content production, divided by the new customers acquired in that period.

Worked: in a month you spend 12,000 dollars on advertising and 8,000 dollars on a fully loaded salesperson, and acquire 40 new customers. CAC is 20,000 divided by 40, or 500 dollars.

Lifetime value is the contribution a customer produces over the whole relationship. For a subscription business the standard approximation is average revenue per user, times gross margin, divided by the churn rate. Suppose average revenue is 60 dollars a month at 80 percent gross margin, giving 48 dollars of monthly contribution, and monthly churn is 5 percent. Average customer life is one divided by 0.05, or 20 months, so lifetime value is 48 times 20, which is 960 dollars.

Two derived numbers matter. The LTV to CAC ratio here is 960 divided by 500, or 1.9. The commonly quoted target of 3 is a heuristic rather than a law, but a ratio below 1 is unambiguous: you are buying revenue at a loss and growth makes it worse. The payback period is CAC divided by monthly contribution, or 500 divided by 48, about 10.4 months. For a cash-constrained business, payback matters more than the ratio, because payback is when the money comes back and can be spent again. A business with a 24 month payback and no outside capital cannot grow, however attractive its lifetime value looks.

Key idea: CAC must be fully loaded, LTV must use gross margin, and payback period is the number that governs whether you can grow without outside money.

How lifetime value gets faked

Lifetime value is the most abused number in entrepreneurship, because it is a forecast wearing the clothing of a measurement. Here are the ways it gets inflated, in rough order of frequency.

• Using revenue instead of gross margin. Multiplying 60 dollars by 20 months gives 1,200 dollars instead of 960, a 25 percent overstatement, and it ignores the cost of delivering the service.
• Using hoped-for churn. Observed churn is 5 percent, but the model assumes it will fall to 2 percent as the product improves. That single substitution raises average life from 20 months to 50 and lifetime value from 960 to 2,400 dollars.
• Averaging only the survivors. Computing average customer tenure from customers who are still subscribed is survivorship bias in its purest form, because the ones who left are exactly the short-tenure customers you excluded.
• Assuming a constant churn rate. The one-divided-by-churn formula assumes the same hazard every month. Real churn is front-loaded: many customers leave in the first ninety days, and the remainder are far stickier. Using a blended early-life churn rate misstates both ends.
• Not discounting. Forty-eight dollars arriving in month nineteen is not worth 48 dollars today, and for a business that is short of cash the difference is not academic.
• Counting expansion that has not happened. Adding assumed upgrades and price increases to a lifetime value that is being used to justify present spending is forecasting your way to a green light.
• Understating CAC. Reporting only paid media while excluding sales salaries, or using blended CAC that includes customers who arrived organically, both flatter the ratio. Blended CAC is a legitimate number, but it is not the number that tells you whether to spend more on advertising.
• Ignoring channel saturation. Your first thousand customers came from the cheapest channel. The next thousand cost more. A CAC measured at small scale rarely holds at ten times the volume.

None of this makes lifetime value useless. It makes it a hypothesis with error bars. State it with the churn rate and margin you actually observed, recompute it every quarter with real cohorts, and treat any version that requires future improvements to work as what it is: a plan, not a measurement.

Key idea: Lifetime value is a forecast, not a measurement. Compute it from observed churn and real gross margin on complete cohorts, and distrust any version that improves whenever it needs to.

Non-subscription businesses

The same logic applies without the subscription vocabulary. For the bakery, lifetime value is visits per year times contribution per visit times years of custom. A neighborhood regular who comes weekly, spends 14 dollars at roughly 65 percent contribution, and stays four years is worth 52 times 9.10 times 4, about 1,893 dollars of contribution. That number should govern how much the bakery is willing to spend to acquire a regular, and it explains why a free loaf to a new neighbor is often a good trade and a coupon in a regional newspaper usually is not.

Try it

A landscaping company charges 180 dollars per monthly visit. Variable costs are 55 dollars of crew time, 20 dollars of fuel and materials, and 5 dollars of processing, so 80 dollars total. Fixed costs are 14,000 dollars a month including the owner's pay. It spends 3,000 dollars a month on advertising and a 2,500 dollar sales commission budget, and adds 22 customers a month. Customers stay an average of 30 months. Compute contribution margin, break-even customers, CAC, LTV, payback, and say what you would change.

Answer: Contribution margin is 180 minus 80, which is 100 dollars per visit, or 55.6 percent. Break-even is 14,000 divided by 100, which is 140 monthly customers. CAC is 5,500 divided by 22, or 250 dollars. LTV is 100 times 30, which is 3,000 dollars, giving an LTV to CAC ratio of 12 and a payback period of 2.5 months. That payback is excellent, which means the correct action is to spend more on acquisition, not less, as long as CAC stays near 250 dollars as volume rises. The thing to watch is whether crews can be hired fast enough to serve the customers, which returns us to the lesson on binding constraints: this business is limited by labor, not by demand or by economics.

Common misconceptions

• "Revenue growth means the model works." A business can grow revenue while losing money on every unit, and growth accelerates the loss.
• "The owner's time is free." Excluding owner pay from fixed costs is the most common way an unviable business appears to break even.
• "A discount will make it up in volume." At a 40 percent margin, a 20 percent discount requires doubling unit sales just to stay even.
• "Low price is a strategy." It is a strategy only with a genuine cost advantage; otherwise it signals low quality and attracts the customers who cost the most to serve.
• "LTV to CAC above three means healthy." Only if LTV uses observed churn and gross margin, and CAC is fully loaded. Payback period usually matters more for a business without outside capital.
• "Payment processing is a rounding error." On low-priced items the fixed per-transaction fee can be close to a tenth of the price.

Recap

• Contribution margin is price minus variable cost; the bakery's is 5.42 dollars on an 8 dollar loaf, or 67.75 percent.
• Break-even units equal fixed costs divided by contribution margin, and adding a 4,000 dollar owner draw moved the bakery from 65 to 94 loaves a day.
• At a 40 percent margin, a 20 percent discount needs a 100 percent volume increase, while a 10 percent price rise survives a 20 percent unit loss.
• Fully loaded CAC includes salaries and tools; LTV uses gross margin and observed churn; payback is CAC divided by monthly contribution.
• Lifetime value is inflated by using revenue, assuming better churn, averaging survivors, ignoring discounting, and understating CAC.
• Non-subscription businesses use visits times contribution times years, which sets a defensible acquisition budget.

Sources

1. U.S. Small Business Administration. (2025). Calculate your startup costs. sba.gov
2. SCORE Association. (2025). Financial projections and break-even templates. score.org
3. Shepherd, D. A., et al. (2020). Financial statements and unit economics. In Entrepreneurship. OpenStax, Rice University. openstax.org
4. Wikipedia contributors. (2025). Customer lifetime value. en.wikipedia.org
5. Wikipedia contributors. (2025). Break-even (economics). en.wikipedia.org

The big picture

The default plan for a first venture is to build the thing. Nine months, most of your savings, a great deal of evening work, and then a launch that answers a question you could have answered in three weeks for four hundred dollars. The point of everything in this lesson is to move the moment of learning as early and as cheaply as possible, because information about whether people want what you are making is the scarcest resource you have.

The vocabulary comes from Eric Ries, whose formulation of the lean startup method organizes the work as a loop: build something, measure how people respond, learn, and go around again. The unit of progress is not features shipped but validated learning, meaning a belief you held that has now been tested against reality. The minimum viable product is whatever gets you around that loop fastest.

What an MVP actually is

The term is misread constantly. Two wrong readings do most of the damage. The first is that an MVP is a bad version of the product, which produces embarrassing launches that teach you only that people dislike broken software. The second is that an MVP is simply the first version you can charge for, which quietly restores the nine-month build.

The useful definition is narrower: an MVP is the cheapest thing that tests the assumption whose failure would most damage the venture. Note what follows from that. The MVP depends on which assumption is riskiest, so two founders in the same market should build different MVPs. And it need not be a product at all. If your riskiest assumption is that dental offices will pay for faster bookkeeping, the MVP is a spreadsheet, an email, and you personally doing four practices' books for a month. No software required.

Key idea: An MVP is not a small product; it is the least expensive test of your most dangerous assumption. Name the assumption first, and the right MVP usually becomes obvious.

Five shapes an MVP can take

Type	What you do	Best for testing
Concierge	Deliver the service by hand to a few customers, openly and manually	Whether the outcome is valuable enough to pay for
Wizard of Oz	Present an apparently working product while humans perform the work behind it	Whether people will use the interface and workflow
Smoke test	A page or ad describing the offer, measuring signups or preorders	Whether the message attracts the intended audience
Prototype	A clickable mockup with no working backend	Whether people understand the product and can complete a task
Preorder or pilot	Take deposits or sell a paid pilot before building	Willingness to pay, the only strong demand signal

Two real cases, with their unglamorous parts intact. Zappos began when Nick Swinmurn photographed shoes in local stores, posted the pictures online, and, when an order came in, went back to the store, bought the shoes at retail, and shipped them. He lost money on every transaction by design. The experiment was not about margin; it was about the single assumption on which the whole business rested, which was whether people would buy shoes they had not tried on. It also required real work and real awkwardness: asking store owners for permission, photographing inventory, handling returns by hand.

Dropbox is the standard smoke test example. The founders made a short video demonstrating a product that did not fully exist, aimed at an audience of technical early adopters, and their waiting list grew enormously overnight. It is worth stating what that story does and does not show. It shows that a credible demonstration to a well-chosen audience can measure interest cheaply. It does not show that any landing page will do this; the video was well made, the audience was specific and reachable, and the founders were technically capable of building what they showed. Retold as "make a landing page and demand will appear," it has misled a great many people.

Key idea: Choose the MVP shape by the risk you need to retire: concierge for value, Wizard of Oz for workflow, smoke test for message, prototype for comprehension, preorder for willingness to pay.

The ethics of testing on real people

Experiments involve actual customers and actual money, so a few duties are not optional.

Do not take money for something you have no realistic ability or intention to deliver. Preselling a product you plan to build is ordinary commerce; preselling one you do not intend to build is fraud, whatever it is called on the pitch deck. If you take deposits, say clearly when delivery is expected, keep the money available to refund, and refund promptly when the date slips or the project stops. In the United States, the Federal Trade Commission's rules on mail, internet, and telephone order merchandise set expectations about shipping within the time promised, or within thirty days when no time is stated, and about giving buyers the option to cancel and receive a refund when you cannot meet it. Crowdfunding platforms add their own obligations. Anything beyond a simple presale is a conversation with a licensed attorney, not a rule of thumb from a course.

Be honest in Wizard of Oz tests about what is happening with customer data and who sees it. "Humans are reading your uploads" is a material fact if your page implies an automated system, and there are contexts, health information most obviously, where that disclosure is a legal requirement rather than a courtesy.

Key idea: Presell only what you intend and are able to build, disclose delivery timing, keep refund money available, and get professional advice before anything more complicated than a simple deposit.

Designing an experiment that can fail

Most founder experiments cannot fail, which is why they teach nothing. The fix is to write four things down before you start.

1. The hypothesis, stated so that it could be false. Not "dentists need better bookkeeping" but "at least 30 percent of two-to-five-chair practices we contact will book a 20 minute call."
2. The metric, chosen in advance, and exactly one of them per hypothesis.
3. The threshold, a number that divides success from failure.
4. The decision you will take at each outcome, written down before you see the data.

A complete example: "We will email 50 independent dental practices from the state association directory. If at least 15 book a 20 minute call within two weeks, we will build the pilot. If between 8 and 14 book, we will run 50 more with a rewritten message. If fewer than 8 book, we will abandon email as a channel and test veterinary referral partnerships instead." That paragraph takes four minutes to write and prevents the most expensive habit in entrepreneurship, which is deciding what the data meant after seeing it.

One arithmetic warning about split tests. Founders love running A and B versions of a page, and most of these tests are underpowered to the point of being decorative. With a baseline conversion of 5 percent, detecting a real improvement to 6 percent, which is a 20 percent relative gain, requires something on the order of eight to nine thousand visitors per version at conventional standards of confidence. With 200 visitors per version you will see about 10 conversions each, and the difference between 10 and 13 is indistinguishable from noise. The practical consequence for a small venture is not to give up on testing; it is to test big changes, where effects are large enough to see, and to stop reading tiny differences as findings.

Key idea: Write hypothesis, metric, threshold, and decision before running anything. Small ventures should test large changes, because small effects require thousands of observations to detect.

Measuring the right things

A vanity metric is a number that rises reliably and tells you nothing you can act on. Cumulative signups, total page views, total registered users, and press mentions are the usual suspects, and their common feature is that they can only go up. An actionable metric is a rate tied to a decision: what share of new signups completed the core action in week one, what share of trial users converted, what a customer costs to acquire in this channel this month.

The most useful discipline here is cohort analysis: group customers by the month they arrived and follow each group separately. Aggregate numbers hide everything. A product whose retention is collapsing can show rising total users for a year as long as new signups exceed departures. Cohorts make the collapse visible immediately, because you can see that the March group behaved worse at day thirty than the January group did.

Watch for the novelty effect as well. Early adopters try things because they are new, and their behavior does not predict the mainstream customer's. A first cohort's enthusiasm is a signal about early adopters only, which is useful information, correctly labeled.

Key idea: Replace cumulative counts with rates, group customers into cohorts by arrival month, and treat early-adopter enthusiasm as evidence about early adopters.

Where this approach does not apply

Experimentation has limits, and pretending otherwise causes harm. A minimum viable version is inappropriate wherever a poor first attempt can injure someone or cannot be undone: medical devices, diagnostics, aviation components, structural engineering, financial custody, children's products. In those fields the regulated development process exists precisely because iterating on live users is unacceptable, and the correct approach is to run the experiments in simulation, bench testing, and formal clinical or engineering studies. Reputation-sensitive categories carry a softer version of the same constraint: a luxury brand or a business selling to a small tight-knit professional community may get exactly one first impression.

And a general caution. Every method in this lesson measures interest, not satisfaction. Email signups and preorders tell you the description was appealing. Only sustained use by paying customers tells you the product was good. Keep those two questions separate in your own head, because conflating them is how a successful smoke test becomes a failed product.

Try it

You believe restaurant owners will pay 300 dollars a month for a service that handles their supplier invoices. Design the cheapest experiment that could prove you wrong, with all four elements written down.

Answer: The riskiest assumption is willingness to pay, not technical feasibility, so the MVP is a concierge test rather than software. Hypothesis: at least 4 of 12 independent restaurants approached will pay 300 dollars for a one month manual pilot in which you personally process their invoices. Metric: signed paid pilots, not expressions of interest. Threshold: 4 of 12. Decision: at 4 or more, run the pilots and measure how many hours each actually takes, which gives you your cost to serve; at 2 or 3, re-run with a different segment such as multi-location operators; at 0 or 1, the willingness to pay is not there at this price and you either test a much lower price point or stop. Note that the pilot also produces the second number you need, since manual delivery reveals your real cost structure before you have written any software.

Common misconceptions

• "An MVP is a cheap version of the product." It is the cheapest test of the riskiest assumption, and it is frequently not a product at all.
• "Ship fast and embarrass yourself." A broken product teaches you that people dislike broken products. The goal is a clean test, not a rushed launch.
• "A landing page proves demand." It measures response to a description. Only payment and sustained use measure demand.
• "A/B testing works at any scale." Detecting a 20 percent relative lift on a 5 percent baseline needs thousands of visitors per version.
• "Growing totals mean things are working." Cumulative counts only rise; cohort retention rates reveal what is actually happening.
• "Every product should be built this way." Regulated, safety-critical, and reputation-critical products require formal development, not iteration on users.

Recap

• The build-measure-learn loop treats validated learning, not shipped features, as the unit of progress.
• An MVP is the least costly test of the assumption whose failure would most damage the venture.
• Concierge, Wizard of Oz, smoke test, prototype, and preorder each retire a different kind of risk.
• Presell only what you intend to build, disclose delivery timing, keep refund money available, and consult a professional beyond simple deposits.
• Write hypothesis, metric, threshold, and decision before running; small ventures should test large changes.
• Use rates and cohorts rather than cumulative totals, and do not apply iteration to safety-critical or regulated products.

Sources

1. Blank, S. (2013). Why the lean start-up changes everything. Harvard Business Review, 91(5), 63-72. hbr.org
2. Wikipedia contributors. (2025). Minimum viable product. en.wikipedia.org
3. Federal Trade Commission. (2025). Business guidance. ftc.gov
4. Shepherd, D. A., et al. (2020). Testing the business idea. In Entrepreneurship. OpenStax, Rice University. openstax.org
5. U.S. Small Business Administration. (2025). Market research and competitive analysis. sba.gov

The big picture

Here is a sentence that surprises people the first time they hear it: profitable businesses go bankrupt regularly, and unprofitable ones survive for years. Profit is an accounting statement about a period. Cash is whether there is money in the account on the fifteenth when payroll runs. They are different quantities, they move at different times, and businesses die of the second one.

The most common shape of this death is not dramatic. A bakery opens, the bread is good, customers come, sales climb every month exactly as the owner hoped, and in month four the account is empty and the equipment loan payment bounces. Nothing went wrong with the business. The owner funded the buildout and forgot to fund the months between opening and break-even. This lesson is about seeing that gap in advance, sizing it, and covering it.

Startup costs, honestly itemized

Start with what it costs to open the doors. Separate one-time costs from ongoing costs, because they are funded differently. For the bakery from the pricing lesson:

One-time cost	Amount
Equipment: deck oven, mixer, proofer, refrigeration	48,000
Leasehold improvements and build-out	35,000
Security deposit (two months rent) and utility deposits	8,100
Signage	3,500
Initial inventory and supplies	3,000
Formation, legal, and accounting setup	2,500
Point of sale and software setup	2,000
Opening marketing	2,000
Licenses, permits, health inspection	1,200
Total one-time	105,300

Most first business plans stop here, and this is the single most expensive omission in small business. The 105,300 dollars opens the door. It does not pay for anything that happens afterward.

The cash-flow gap, worked month by month

Recall the bakery's economics: contribution margin of 67.75 percent, and fixed costs of 13,200 dollars a month including a 4,000 dollar owner draw. Break-even is about 19,500 dollars of monthly sales. No new bakery does 19,500 dollars in month one. Here is a realistic ramp.

Month	Sales	Contribution at 67.75%	Monthly cash flow	Cumulative
1	8,000	5,420	-7,780	-7,780
2	11,000	7,453	-5,747	-13,527
3	14,000	9,485	-3,715	-17,242
4	17,000	11,518	-1,682	-18,925
5	19,500	13,211	+11	-18,914
6	22,000	14,905	+1,705	-17,209
7	24,000	16,260	+3,060	-14,149

The peak cash requirement is about 18,900 dollars, reached at the end of month four. That is the working capital the owner must have on hand in addition to the 105,300 dollars of one-time costs, or the business fails while succeeding. Total realistic capital need is therefore about 124,000 dollars, not 105,000.

Now the part that separates a plan from a fantasy. Run the ramp 25 percent slower, which is a modest and extremely common miss. Month one becomes 6,000 dollars instead of 8,000, and break-even arrives in month eight rather than month five. Recompute the cumulative deficit and the peak cash need rises to about 33,700 dollars: a 25 percent slower ramp nearly doubles the cash requirement. This is why experienced lenders and advisers add a reserve on top of the calculated gap rather than funding it exactly. A reserve of three to six months of fixed costs, so 40,000 to 79,000 dollars here, is not timidity; it is an accurate reading of how forecasts behave.

Key idea: Total capital need equals one-time costs plus the peak cumulative cash deficit plus a reserve. A modestly slower revenue ramp can double the peak deficit, so the reserve is part of the plan, not padding.

Why growth itself consumes cash

There is a second cash trap that hits businesses which are working well. If you invoice customers at net thirty and they actually pay in forty-five days, while you pay staff every two weeks and suppliers in thirty, then every new client is funded out of your pocket for about six weeks before their money arrives. Serving twice as many clients means carrying twice as much of that float. A commercial cleaning company growing 20 percent a month can be perfectly profitable on every contract and still run out of money, because the growth is consuming working capital faster than the profits replace it.

The levers are the ones you would guess: invoice immediately rather than at month end, take deposits or retainers up front, offer a small discount for prepayment, ask suppliers for longer terms, and refuse the customer who insists on paying in ninety days unless the margin genuinely pays for the float. Module 6 works this arithmetic in full. For now, note the principle: growth is an expense before it is a benefit.

Bootstrapping

Bootstrapping means funding the business from personal savings and from its own revenue rather than from outside investors. It is not a fringe strategy; it is what most American businesses do. Federal Reserve surveys of small employer firms consistently find that personal savings and owner funds are the largest source of startup capital, well ahead of every institutional option.

The tactics that make it work all point in the same direction, which is converting fixed costs into variable ones and pulling cash forward:

• Pre-sell. Deposits, retainers, annual prepayment at a discount, and paid pilots are customer-funded capital that costs no equity and no interest.
• Invoice on milestones rather than on completion, and invoice the day the milestone is met.
• Use contractors before employees, and rented or used equipment before purchased new equipment, until volume is proven.
• Run a service business that funds a product business. A great many software companies were paid for by consulting.
• Start while employed. Most businesses begin as side projects, and the paycheck is the cheapest capital available.
• Negotiate supplier terms deliberately. Thirty extra days from a supplier is an interest-free loan you did not have to apply for.

Bootstrapping also has real costs, and the culture around it tends to treat it as a virtue rather than a trade-off. It is slower. It concentrates all the risk on the founder's household. It can cost you a market where scale matters and a funded competitor gets there first. And undercapitalization is itself among the leading causes of small business failure, so a founder who bootstraps into a business that never had enough cash to reach break-even has not been prudent, only underfunded. Choose it because the arithmetic supports it, not because taking money feels like weakness.

Key idea: Bootstrapping is the majority path and works by pulling cash forward and keeping costs variable, but it trades speed for control and can shade into undercapitalization, which is itself a common cause of failure.

Friends, family, and the money that costs the most

After personal savings, money from people who know you is the most common outside capital. The financial risk is ordinary; the relational risk is not. Four rules make it survivable.

Write it down. A one-page note stating amount, terms, and what happens if the business fails is worth more than any amount of goodwill, because memory reconstructs favorable versions of undocumented conversations. Say the base rate out loud: most new businesses do not survive ten years, and this money may be entirely lost. Never accept money someone needs, for retirement, for a child's tuition, for medical costs, however much they want to help. And decide deliberately between a loan and equity, because they behave very differently when things go badly.

One legal point that founders regularly miss: selling an ownership stake is selling a security, and that is regulated by federal and state law regardless of whether the buyer is your uncle. Exemptions exist for small private offerings, and they have conditions. This is genuinely a matter for a licensed securities attorney before you accept the check, not after. The same caution applies to promising anyone a share of profits.

Debt, including what the SBA actually does

Debt is cheaper than equity when you can service it, because interest is finite and equity is forever. It is also unforgiving: the payment is due whether or not the month went well.

A common misunderstanding is worth correcting. The Small Business Administration does not, outside of disaster loans, lend money. It guarantees a portion of loans made by banks and other participating lenders, which reduces the lender's risk and makes credit available to borrowers who would otherwise be declined. The main programs are the 7(a) program for general purposes up to five million dollars, the 504 program for fixed assets such as real estate and heavy equipment through Certified Development Companies, and microloans of up to 50,000 dollars made through nonprofit intermediaries, where the typical loan in practice is far smaller, often in the low tens of thousands.

Two realities to plan around. First, lenders generally want operating history, usually two years or more, plus collateral and a solid personal credit score, which means conventional debt is frequently unavailable at the true startup stage, exactly when you want it. Second, small business loans almost always require a personal guarantee. That means the limited liability of your LLC or corporation does not protect you from this debt. If the business fails, the lender can pursue your personal assets. Founders who form an LLC believing it makes them immune to business debts are usually wrong about the one debt that matters most.

Key idea: The SBA guarantees loans rather than making them, lenders want history and collateral, and a personal guarantee punches straight through your limited liability.

Crowdfunding, in two very different forms

Reward-based crowdfunding on platforms like Kickstarter and Indiegogo is presales with a marketing campaign attached. Money arrives before you build, which is excellent, and it is not free: platform and payment fees commonly run in the high single digits, you owe every backer a product, and the record of funded hardware projects delivering on time is poor. Treat a campaign as a demand test plus working capital plus a delivery obligation, and budget the fulfillment costs, which founders routinely underestimate.

Equity crowdfunding under Regulation Crowdfunding is a different animal: you are selling securities to the general public through a registered portal, subject to annual limits, disclosure requirements, and ongoing reporting. It has opened real access to capital, and it comes with legal obligations that make it professional territory from the first step. Talk to a securities attorney before you begin, not when something goes wrong.

Try it

A landscaping startup has one-time costs of 62,000 dollars for a truck, trailer, and equipment. Fixed costs are 9,000 dollars a month including the owner's pay, and contribution margin is 55 percent. Sales ramp: 6,000, 9,000, 12,000, 15,000, 17,000, then steady. Find break-even, the peak cash requirement, and total capital needed.

Answer: Break-even revenue is 9,000 divided by 0.55, which is about 16,364 dollars a month, reached in month five. Monthly contributions are 3,300, 4,950, 6,600, 8,250, and 9,350, so monthly cash flows are -5,700, -4,050, -2,400, -750, and +350. The cumulative deficit runs -5,700, -9,750, -12,150, -12,900, then -12,550, so the peak cash requirement is about 12,900 dollars at the end of month four. Total capital needed is 62,000 plus 12,900, or roughly 75,000 dollars, and a prudent plan adds three months of fixed costs, about 27,000 dollars, bringing the honest figure closer to 102,000. Note that the equipment is collateral, which makes this business a plausible candidate for a secured loan or an SBA-guaranteed one, unlike the software business in the next lesson.

Common misconceptions

• "If the business is profitable it will survive." Profit is an accounting result for a period; insolvency is about cash on a specific day.
• "Startup costs mean the cost of opening." They also include the working capital that covers losses until break-even, which is the commonly omitted half.
• "Growth solves cash problems." Growth consumes working capital first and returns it later, which is why fast-growing profitable firms run out of money.
• "The SBA lends money to startups." Except for disaster loans, it guarantees loans made by private lenders, who still apply their own credit standards.
• "An LLC protects me from business debt." Not from debt you personally guaranteed, and small business lenders nearly always require a personal guarantee.
• "Money from family is informal." Selling equity is selling a security regardless of the relationship, and the rules apply to your uncle's check too.

Recap

• Total capital need equals one-time costs plus the peak cumulative cash deficit plus a reserve of three to six months of fixed costs.
• The bakery needed about 105,300 dollars to open and about 18,900 more to reach break-even, and a 25 percent slower ramp raised the gap to about 33,700.
• Growth consumes cash through receivables and payroll timing before it produces any.
• Bootstrapping is the majority path, works by pulling cash forward and keeping costs variable, and risks undercapitalization if pushed too far.
• Friends and family money requires written terms, an explicit statement of the base rate, and attention to securities law.
• The SBA guarantees rather than lends, lenders want history and collateral, and personal guarantees defeat limited liability.

Sources

1. U.S. Small Business Administration. (2025). Fund your business. sba.gov
2. U.S. Small Business Administration. (2025). Loans. sba.gov
3. Federal Reserve Banks. (2025). Small Business Credit Survey. fedsmallbusiness.org
4. U.S. Securities and Exchange Commission. (2025). Investor bulletins on crowdfunding and private offerings. Investor.gov. investor.gov
5. SCORE Association. (2025). Startup cost and cash flow templates. score.org

The big picture

Venture capital is discussed as though it were a prize awarded for excellence. It is not. It is a specific financial product, designed for a specific and unusual kind of company, sold on terms that are entirely rational for the buyer and frequently misunderstood by the seller. Almost everything that confuses founders about venture capital dissolves once you understand what a fund is obliged to do with its money.

This lesson explains the mechanics honestly, works a real cap table with arithmetic you can reproduce, and then answers the question that actually matters, which is whether this product fits your business. For most readers of this course the answer will be no, and no is a perfectly good answer that leaves you with a perfectly good company.

One boundary before we start. Nothing here is investment advice, and nothing here should be applied to your own term sheet without a startup attorney. Financing documents are the place where a small misunderstanding costs the most, and the cost is measured in years of your life.

What a venture fund has to do

A venture capital firm raises a fund from limited partners: pension funds, university endowments, foundations, insurers, wealthy families. The fund has a defined life, conventionally around ten years. The firm typically charges an annual management fee of about two percent of committed capital and keeps roughly twenty percent of the profits, an arrangement called carried interest. To be considered successful and to raise another fund, the firm needs to return the entire fund several times over.

Now put that together with the return distribution from Module 1. Roughly 65 percent of venture investments return less than the capital put in, and about four percent return more than ten times. A fund with thirty investments should expect most of them to produce nothing. The only way the arithmetic works is if one or two investments return an amount comparable to the whole fund.

That single fact explains nearly every venture behavior founders find puzzling. It explains why an investor is uninterested in a company that will reliably earn four million dollars a year forever: a two hundred million dollar fund cannot be returned by any share of that. It explains why investors push for aggressive growth even at the cost of profitability, since a company that grows slowly to a moderate size is, from the fund's perspective, indistinguishable from a failure. And it explains why they will encourage a risky strategy that might produce an enormous outcome over a safe one that produces a good one. The fund is diversified across thirty draws from a power law. You are one draw. Your interests genuinely diverge, and pretending otherwise helps nobody.

Key idea: A venture fund needs individual investments capable of returning a large fraction of the entire fund, which restricts it to companies with a plausible path to a very large outcome and makes it a poor fit for most good businesses.

Angels and the instruments

Before funds, there are individuals. Angel investors are people investing their own money, commonly in amounts from ten thousand to a hundred thousand dollars, often former operators in the industry, frequently organized into groups that pool diligence and checks. In the United States most private offerings are made to accredited investors, a category the Securities and Exchange Commission defines mainly through income or net worth thresholds, with additional routes through certain professional licenses. Angel returns follow the same power law as funds, and most individual angels lose money; the ones who do well are usually those who made many small investments rather than a few large ones.

Three instruments dominate early rounds:

• Priced equity round. You agree a valuation and sell preferred stock. Most expensive to paper, and the cleanest to understand, because everyone knows exactly what they own.
• Convertible note. A loan that converts into equity at the next priced round, usually with a discount and a valuation cap, plus interest and a maturity date. The maturity date is real: if the round does not happen, you owe money.
• SAFE. A simple agreement for future equity. Not debt, no interest, no maturity, converting at the next priced round subject to a cap or discount. Fast and cheap, and with a trap: multiple SAFEs stack, and founders who raise several before any priced round frequently discover they have sold much more of the company than they believed. Model the conversion of every outstanding instrument before signing the next one.

A cap table, worked

Two founders incorporate and split ownership evenly: 8,000,000 shares, 4,000,000 each. Here is what three rounds do.

Seed. An investor puts in 1,500,000 dollars at a 6,000,000 dollar pre-money valuation, making the post-money valuation 7,500,000 dollars, so the investor takes 20 percent. The investor also requires a 10 percent employee option pool, created before the round closes. Post-round ownership must therefore be: investor 20 percent, pool 10 percent, founders 70 percent.

Work the shares. The founders' 8,000,000 shares represent 70 percent, so total shares become 8,000,000 divided by 0.70, which is 11,428,571. The investor receives 20 percent of that, 2,285,714 shares, and the pool holds 1,142,857. Price per share is 1,500,000 divided by 2,285,714, which is about 0.65625 dollars.

Here is the detail worth learning, sometimes called the option pool shuffle. At that share price, the founders' existing 8,000,000 shares are worth 5,250,000 dollars, not the 6,000,000 dollar pre-money valuation that was negotiated. The pool was carved out of the pre-money, which means the founders paid for all of it. This is standard practice rather than a trick, and it is negotiable, and a great many founders sign without noticing that the headline valuation overstates what they received by three quarters of a million dollars.

Series A and B. Suppose each subsequent round sells 20 percent of the company: 6,000,000 dollars at a 24,000,000 pre-money, then 15,000,000 dollars at a 60,000,000 pre-money. Each round multiplies every existing holder's percentage by 0.80.

Holder	At founding	After seed	After Series A	After Series B
Founder A	50.0%	35.0%	28.0%	22.4%
Founder B	50.0%	35.0%	28.0%	22.4%
Option pool	-	10.0%	8.0%	6.4%
Seed investor	-	20.0%	16.0%	12.8%
Series A	-	-	20.0%	16.0%
Series B	-	-	-	20.0%

Each founder now owns 22.4 percent, having raised 22,500,000 dollars in total. The standard and correct defense of this arithmetic is that 22.4 percent of a large company beats 50 percent of a small one. That defense is true whenever the capital actually caused the growth. It is false whenever the capital simply funded a larger version of a business that was not working, which is why the diligence you do on your own model before raising matters more than the terms you negotiate.

Key idea: Three rounds at 20 percent each take two 50/50 founders to 22.4 percent apiece, and an option pool created pre-money is paid for entirely out of the founders' side of the table.

Liquidation preference, and why the percentage lies

Investors buy preferred stock, and preferred stock gets paid first. A 1x non-participating liquidation preference, the most common form, means each investor receives the greater of their money back or their converted pro-rata share of the proceeds. That sounds mild. Watch what it does.

At a 150,000,000 dollar sale, every series does better converting to common, so each founder receives 22.4 percent of 150,000,000, which is 33,600,000 dollars. The percentages behave exactly as the table suggests.

At a 40,000,000 dollar sale, they do not. Series B invested 15,000,000 for 20 percent; 20 percent of 40,000,000 is only 8,000,000, so Series B takes its 15,000,000 preference instead. That leaves 25,000,000 for everyone else, and now Series A faces the same choice: its share of the remainder falls below its 6,000,000 preference, so it takes the preference too. After 21,000,000 of preferences, 19,000,000 remains, divided among the founders, the seed investor, and the pool in proportion to their shares. The founders together hold 44.8 percent of a remaining 64 percent, so they receive 70 percent of 19,000,000, which is 13,300,000 dollars, or about 6,650,000 dollars each.

The naive calculation, 22.4 percent of 40,000,000, would have predicted 8,960,000 dollars each. The preference stack cost each founder roughly 2,300,000 dollars, and the calculation had to be worked iteratively because each series chooses whichever option pays it more. At exits below the total preference stack, founders and employees can receive nothing at all while investors are made whole. Employees holding options in that situation frequently learn about this structure for the first time on the day of the sale.

Key idea: Ownership percentage describes proceeds only above the preference stack. Below it, preferred holders are paid first, and founders and employees can receive far less than their percentage implies, or nothing.

The other terms that matter

Valuation is the number founders negotiate hardest and rarely the term that hurts most. Ask an attorney about all of these: board composition, which determines who can fire you; protective provisions, which give investors a veto over sales, new financings, and budgets; anti-dilution protection, where a broad-based weighted average formula is standard and a full ratchet is punishing in a down round; pro-rata rights to participate in future rounds; drag-along rights, which can force you to sell; and participating preferred, which lets an investor take its money back and its share of the remainder. A clean term sheet at a lower valuation is usually better than an aggressive valuation loaded with structure.

When venture capital actually fits

It fits when a large amount of capital deployed now creates a durable position that cannot be built incrementally. Concretely: markets with network effects where the leader's advantage compounds; long research and development cycles before any revenue, as in therapeutics or semiconductors; genuinely capital-intensive buildouts; and winner-take-most categories where being third is worth very little.

It does not fit consultancies, agencies, most local service businesses, most restaurants and retail, or software businesses that would happily reach four million dollars of revenue and stay there. None of those are lesser businesses. They are simply the wrong shape for a product built to chase outliers. And a final honest note: taking venture money changes what counts as success. The comfortable, profitable, five-million-dollar company you would have been delighted to own becomes, on the fund's books, a failure. Decide whether you want that objective function before you sign, because it does not come off afterward.

Key idea: Venture capital fits network-effect, deep-research, and capital-intensive businesses aiming at very large outcomes. Accepting it replaces your definition of success with the fund's.

Try it

A founder owns 100 percent of 5,000,000 shares. She raises 2,000,000 dollars at an 8,000,000 dollar pre-money valuation with a 15 percent post-money option pool created pre-money. What percentage does she hold after the round, and what is her effective per-share price?

Answer: Post-money is 10,000,000 dollars, so the investor takes 20 percent. With a 15 percent pool, the founder retains 65 percent. Her 5,000,000 shares therefore represent 65 percent of the company, so total shares become 5,000,000 divided by 0.65, which is 7,692,308. The investor receives 20 percent, or 1,538,462 shares, at a price of 2,000,000 divided by 1,538,462, which is 1.30 dollars per share. Her stake is worth 5,000,000 times 1.30, which is 6,500,000 dollars, not the 8,000,000 pre-money headline. The 1,500,000 dollar difference is the option pool, and she paid for all of it.

Common misconceptions

• "Raising venture capital is a milestone of success." It is the purchase of a financial product, and it is unsuitable for the large majority of viable businesses.
• "Investors want steady profitable growth." A fund needs outlier outcomes; a reliably profitable moderate business cannot return a fund.
• "Valuation is the most important term." Board control, protective provisions, anti-dilution, and preference structure routinely matter more.
• "My percentage tells me what I will receive." Only above the preference stack. Below it, preferred investors are paid first and common holders may receive nothing.
• "SAFEs are too simple to hurt anyone." Multiple SAFEs stack, and their combined conversion regularly surprises founders at the next priced round.
• "The option pool comes out of everyone." A pool created pre-money is funded entirely by existing holders, which at the seed stage means the founders.

Recap

• Venture funds need investments capable of returning a large fraction of the whole fund, which is why they pursue outliers.
• Angels invest personally, usually as accredited investors, and their returns follow the same power law.
• Priced rounds, convertible notes, and SAFEs differ in maturity, interest, and how surprises accumulate.
• Three 20 percent rounds took two 50/50 founders to 22.4 percent each, and the pre-money option pool cost them 750,000 dollars of headline value at seed.
• At a 40,000,000 dollar exit the preference stack reduced each founder from a naive 8,960,000 dollars to about 6,650,000.
• Venture capital suits network-effect, deep-research, and capital-intensive ventures, and it replaces your definition of success with the fund's.

Sources

1. U.S. Securities and Exchange Commission. (2025). Accredited investors and private offerings. Investor.gov. investor.gov
2. National Venture Capital Association. (2025). Venture industry data and model legal documents. nvca.org
3. Wikipedia contributors. (2025). Venture capital. en.wikipedia.org
4. Wikipedia contributors. (2025). Capitalization table. en.wikipedia.org
5. Ewing Marion Kauffman Foundation. (2025). Entrepreneurship and capital access research. kauffman.org

The big picture

Open almost any first business plan and find the same chart. Revenue is modest in year one, doubles in year two, and by year five reaches a number in the tens of millions. The curve bends upward at exactly the point where the author's actual knowledge ends. Everyone who reads business plans knows this chart is fiction. Frequently the person who drew it knows too, which raises a fair question: why produce projections at all?

Because the useful part was never the prediction. A projection is a machine for finding out three things: how much cash you need and when, which assumptions actually drive the outcome, and whether the business has a ceiling you have not noticed. Those three answers are available before you spend any money, and they are worth having even though the revenue line will be wrong.

What the statements do

There are three financial statements and they answer different questions. The income statement reports revenue and expenses over a period and produces profit. The balance sheet reports what you own and owe at a moment. The cash flow statement reports money actually moving in and out.

For a new venture the cash flow statement is the one that matters most, for the reason established two lessons ago: businesses die of cash, not of accounting losses. The gap between the two is timing. Under accrual accounting you record revenue when it is earned, so an invoice sent in March is March revenue even if the customer pays in May. That is the correct way to measure performance and a dangerous way to manage a bank account. Small businesses may often use cash-basis accounting instead, and the choice has tax consequences; which method you may use and should use is a question for a CPA, not for a course.

Key idea: Projections exist to size the cash requirement, expose the driving assumptions, and reveal ceilings. For a new venture, the cash flow statement outranks the income statement.

Driver-based forecasting, worked

A top-down forecast says the market is two billion dollars and you will take two percent by year three. It is unfalsifiable and useless. A driver-based forecast builds revenue from operational quantities you can observe and manage.

Take a business selling software to small firms at 400 dollars a month. The drivers are: two salespeople, each holding 40 qualified conversations a month, closing 15 percent of them. That is 80 conversations, 12 new customers a month, and 4,800 dollars of new monthly recurring revenue added each month. Monthly churn is 3 percent of the existing base.

Now roll it forward. Each month's ending recurring revenue is the prior month times 0.97, plus 4,800.

Month	Monthly recurring revenue	Month	Monthly recurring revenue
1	4,800	7	30,722
2	9,456	8	34,600
3	13,972	9	38,362
4	18,353	10	42,011
5	22,602	11	45,551
6	26,724	12	48,984

Year one revenue is the sum of those twelve figures, about 336,000 dollars, exiting at roughly 49,000 dollars a month, which annualizes to about 588,000 dollars. Notice how different that is from the intuitive answer. A founder who reasoned "twelve customers a month times 400 dollars times twelve months" would have projected 691,200 dollars, because that calculation silently assumes every customer arrives in January and nobody ever leaves.

The ceiling nobody draws

Now the most useful thing this model will tell you. If you keep adding 4,800 dollars of new recurring revenue every month and lose 3 percent of the base every month, the two forces eventually balance. Losses equal additions when 0.03 times the base equals 4,800, so the base converges to 4,800 divided by 0.03, which is 160,000 dollars a month, or 1.92 million dollars a year. Not in year five. Ever, at this sales capacity and this churn rate.

That single number does more work than any five-year chart. It says the plan has a hard ceiling, and it says exactly which two levers move it. Watch what churn does:

Monthly churn	Steady-state MRR	Implied annual revenue
1.5 percent	320,000	3.84 million
3 percent	160,000	1.92 million
5 percent	96,000	1.15 million
8 percent	60,000	0.72 million

Halving churn doubles the entire company. No amount of additional selling does that, because selling raises the numerator linearly while churn divides it. A founder who understands this stops asking for a bigger marketing budget and starts asking why customers leave in month four. This is what a projection is for.

Key idea: Constant additions against a percentage churn converge to additions divided by churn. That steady state is the real size of the business, and reducing churn moves it far more than selling harder.

Why hockey sticks are not credible

The characteristic chart fails for specific, diagnosable reasons. Each one is worth being able to name.

• No mechanism for the inflection. The curve bends and nothing in the model bends with it. If revenue triples in year three, something must triple: reps, channels, price, or conversion. Name it or delete the bend.
• Costs held flat while revenue multiplies. Ten times the revenue with the same two salespeople is not a forecast, it is an arithmetic error. Costs must be driven by the same quantities that drive revenue.
• Implied share that exceeds reality. Carry the growth rate forward two more years than the chart shows. If year seven exceeds the entire market, the curve was never a model.
• Ignoring the steady state. Any model with churn has a ceiling. A projection that grows through it has simply not been computed properly.
• The word conservative attached to the aggressive case. If the plan labelled conservative is also the plan you are counting on, you have no downside case at all.
• The planning fallacy. People forecasting their own projects systematically underestimate time and cost, and the correction is not more optimism-checking but reference class forecasting: find what actually happened to a set of comparable ventures and start from their distribution rather than from your plan.

Key idea: A credible projection names the mechanism behind every change in slope, scales costs with the same drivers as revenue, respects the steady state, and starts from what happened to comparable ventures.

Costs that people forget

Revenue lines get all the attention and cost lines cause most of the damage. Four recurring omissions:

Fully loaded payroll. An employee costs materially more than their salary. Employer payroll taxes, unemployment insurance, workers compensation, benefits, equipment, and software commonly add twenty to thirty percent or more on top of base pay. A plan that budgets 60,000 dollars for a 60,000 dollar salary is short by roughly 15,000 dollars a year per person.

Your own compensation. Same point as the break-even lesson, and worth repeating because it is the most common single error in small business planning.

Working capital. Growth ties up cash in receivables and inventory. A projection showing profit while cash falls is not necessarily wrong; it may be correctly showing you a growth-driven cash drain.

Taxes, and the miscellaneous bucket. A line reading "other expenses, 10 percent of revenue" tells a reader you have not costed the business. Itemize it. And do not model your own tax position from a spreadsheet template: entity type, self-employment tax, estimated payments, and deductions interact in ways that genuinely require a CPA.

Three scenarios and the only question that matters

Build three cases, and make them differ in drivers rather than in a fudge percentage. The base case is what you actually expect. The downside holds costs where they are and cuts the drivers hard: close rate down a third, ramp two months slower, churn a point higher. The upside raises the drivers you have some evidence for.

Then ask the question the whole exercise exists to answer: in the downside case, does the business survive, and for how many months? Runway is cash on hand divided by monthly net burn. If the downside case runs out of money in month seven and your sales cycle is four months, you do not have a plan, you have a hope. The fix is one of three things: raise more, cut fixed costs so burn falls, or change the model so revenue arrives earlier through deposits and prepayment. Choosing among those three is the actual work of financial planning, and it is impossible without having built the scenarios first.

Key idea: Scenarios must differ in drivers, not in a blanket percentage, and the decisive output is whether the downside case survives long enough for the business to work.

Try it

A founder projects year one revenue of 300,000 dollars, year two of 1.5 million, and year three of 7 million, with headcount rising from 3 to 5 to 7. Identify the defects and say what you would ask for.

Answer: Revenue rises roughly 23 times over two years while headcount slightly more than doubles, so revenue per employee climbs from 100,000 dollars to one million. That is possible only in a business with near-zero marginal cost and a self-serve channel, and the plan should state which. Nothing in the model explains the inflection: no new channel, no price change, no conversion improvement, so the growth is asserted rather than derived. There is no churn assumption, which means no steady state has been computed. And the headcount is not tied to the drivers, so nobody has asked who will serve the customers implied by seven million dollars. What to ask for: the four drivers behind each year's revenue, the churn rate, the steady state those imply, and a downside case with the same cost structure.

Common misconceptions

• "Projections are supposed to be accurate." They are supposed to be derived. Readers judge whether you understand your own drivers, not whether you can see the future.
• "Profit means solvency." Cash and profit diverge through timing, and only the cash statement tells you whether payroll clears.
• "More sales fixes everything." With churn, the steady state is additions divided by churn, so retention often moves the ceiling more than acquisition does.
• "My projections are conservative." If the conservative case is the one you are counting on, no downside case exists.
• "Salary is the cost of an employee." Payroll taxes, insurance, benefits, and equipment typically add twenty to thirty percent or more.
• "Growth always improves cash." Growth consumes working capital first; a profitable, fast-growing firm can be the one that runs out of money.

Recap

• Projections size the cash requirement, expose driving assumptions, and reveal ceilings; the cash flow statement matters most early.
• Driver-based forecasting builds revenue from reps, conversations, close rates, and price rather than from market share assumptions.
• The worked model reached about 336,000 dollars in year one and exited at roughly 49,000 dollars of monthly recurring revenue.
• Constant additions against percentage churn converge to additions divided by churn; at 3 percent that ceiling was 160,000 dollars a month.
• Hockey sticks fail by omitting a mechanism, holding costs flat, ignoring the steady state, and mislabeling the aggressive case as conservative.
• Scenarios should differ in drivers, and the decisive test is whether the downside case leaves enough runway.

Sources

1. U.S. Small Business Administration. (2025). Write your business plan. sba.gov
2. SCORE Association. (2025). Financial projections templates and startup financial statements. score.org
3. Internal Revenue Service. (2025). Small businesses and self-employed: accounting periods and methods. irs.gov
4. Wikipedia contributors. (2025). Planning fallacy. en.wikipedia.org
5. Shepherd, D. A., et al. (2020). Entrepreneurial finance and accounting. In Entrepreneurship. OpenStax, Rice University. openstax.org

The big picture

Start with the boundary, because this is the lesson where it matters most. Nothing here is legal or tax advice. Entity choice, tax elections, and intellectual property filings depend on facts about you: which state you are in, whether you have partners, how you will be paid, whether you plan to raise money, what you have already disclosed publicly. The purpose of this lesson is to give you the vocabulary and the map so that an hour with a licensed attorney and an hour with a CPA are productive rather than remedial. Those two hours are among the highest-return money a founder spends, and this course cannot substitute for them.

What a course can do is tell you which decisions have deadlines, which are cheap to reverse, and which are not. That is most of the value, because founders rarely go wrong by choosing the wrong entity. They go wrong by missing a filing window that never reopens.

The five structures

Structure	Liability	Federal taxation	Typical fit
Sole proprietorship	None; owner and business are one	Schedule C on personal return; self-employment tax on net earnings	Solo, low-risk, testing an idea
General partnership	None; partners jointly and severally liable	Partnership return with K-1s to partners	Rarely chosen deliberately; often the accidental default
LLC	Limited, if formalities are respected	Default: disregarded entity or partnership; may elect corporate or S treatment	Most small businesses with real liability exposure
S corporation	Depends on the underlying entity	Pass-through; owner takes reasonable W-2 wages plus distributions	Profitable owner-operated firms where the payroll tax saving exceeds the added complexity
C corporation	Limited	Separate taxpayer at the federal corporate rate; dividends taxed again	Ventures raising institutional capital, multiple share classes

Three clarifications that resolve most confusion. First, a sole proprietorship is a default, not a choice: if you start selling and do nothing else, that is what you are. Second, an S corporation is not an entity type; it is a federal tax election that an eligible corporation or LLC makes. Eligibility is restrictive: broadly, no more than one hundred shareholders, only individuals and certain trusts and estates, generally United States citizens or residents, and a single class of stock. The attraction is that after paying yourself reasonable compensation as wages, remaining profits are distributed without self-employment tax. The IRS takes the reasonable compensation requirement seriously, and setting your own salary too low to avoid payroll tax is a well-known audit issue. Whether the election helps you at your profit level is exactly a CPA question. Third, venture investors will generally require a C corporation, conventionally a Delaware one, because they need preferred stock, multiple classes, and a structure their own investors can hold. If you intend to raise institutional money, converting later is possible but costs time and legal fees.

Key idea: The LLC is the default answer for most small businesses with liability exposure, the S election is a tax question for profitable owner-operators, and the C corporation is the price of admission for institutional investment.

What limited liability does not do

Founders routinely overestimate the shield. Limited liability separates the business's debts and many of its claims from your personal assets. It does not protect you from four things.

• Your own conduct. If you personally act negligently or wrongfully, you can be liable regardless of the entity. The company being sued does not make you unsuable.
• Debts you personally guaranteed. As the funding lesson established, small business lenders and many commercial landlords require personal guarantees, which reach straight through the entity.
• Certain tax obligations. Unpaid payroll taxes withheld from employees can be recovered personally from responsible persons. Using withheld payroll tax to cover a cash shortfall is one of the most dangerous things a struggling founder can do.
• A pierced veil. Courts can disregard the entity when owners treat it as an alter ego. The recurring facts are commingling personal and business funds, skipping required formalities and records, gross undercapitalization at formation, and fraud.

The practical consequences are cheap and boring: a separate business bank account from day one, no personal expenses run through it, minutes or written consents where your state expects them, an operating agreement even for a single-member LLC, and adequate insurance. Insurance is the protection founders skip and lawyers mention first: general liability, professional liability where you give advice, product liability where you make things, and workers compensation where your state requires it.

Key idea: Limited liability fails against your own acts, personal guarantees, withheld payroll taxes, and veil piercing. Separate accounts, real records, and insurance are what actually protect a founder.

Getting set up

An employer identification number comes free directly from the IRS, and a large number of paid services exist that charge for the same form; you do not need them. Beyond that: register the entity with your state, appoint a registered agent, check city and county licensing, register for sales tax if you sell taxable goods or services, and confirm any industry-specific permits, which are where food, health, childcare, construction, and transportation businesses spend real time. None of this is intellectually difficult. All of it takes longer than you expect, which is why it belongs on the timeline before your planned opening date rather than after.

Intellectual property, four different animals

Patents protect inventions. A utility patent generally runs twenty years from the filing date; a design patent, covering ornamental appearance, runs fifteen years from grant for recent filings. Since the America Invents Act the United States awards patents on a first-inventor-to-file basis, which makes filing dates decisive. A provisional application is a lower-cost filing that establishes a date and permits the phrase patent pending for twelve months, after which a non-provisional must be filed or the priority is lost. That twelve-month deadline is unforgiving and it is a calendar item, not a judgment call.

The disclosure rule catches more founders than any other. The United States allows a one-year grace period after your own public disclosure, but most other countries apply absolute novelty: publishing, demonstrating publicly, or offering the invention for sale before filing destroys your foreign patent rights permanently. A crowdfunding launch, a conference demo, or a detailed blog post can be a disclosure. If you have an invention worth protecting internationally, talk to a patent attorney before you show it to anyone.

Be realistic about cost and enforcement. A provisional prepared with counsel typically runs a few thousand dollars, and a utility patent through issuance commonly reaches five figures once attorney time and office actions are included, though the USPTO offers reduced fees for small and micro entities. And a patent is not self-enforcing: it is a right to sue, exercised at your expense, and patent litigation costs are far beyond what most small ventures can carry. For many businesses that reality makes speed and customer relationships a better moat than a filing.

Trademarks protect the marks that identify the source of goods and services. In the United States rights arise from actual use in commerce, and federal registration with the USPTO adds nationwide priority, the right to use the registered symbol, and much easier enforcement. Strength depends on distinctiveness: fanciful invented words are strongest, then arbitrary words unrelated to the product, then suggestive marks, then descriptive marks which are protectable only after acquiring secondary meaning, and finally generic terms which can never be protected. The practical lesson is to run a clearance search before you commit to a name, and certainly before you print signage, buy the domain, and wrap a van. Renaming a business with existing customers is expensive in a way founders always underestimate.

Copyright protects original expression fixed in a tangible medium: code, text, photographs, designs, music. It exists automatically at creation with no registration required. Registration still matters, because for United States works it is generally a prerequisite to filing an infringement suit and to recovering statutory damages and attorney fees.

Here is the trap that catches nearly every first-time founder. Work made for hire covers employees acting within the scope of employment. It generally does not cover independent contractors. So the freelance designer who made your logo, the developer who built your first app, and the photographer who shot your product images may own the copyright in that work even though you paid for it, unless you have a signed written assignment. Put an intellectual property assignment clause in every contractor agreement, before the work starts. Fixing this after the fact requires the contractor's cooperation, which you may not have.

Trade secrets protect commercially valuable information that you keep secret through reasonable measures: formulas, processes, customer lists, pricing models, source code. There is no registration and no expiration; protection lasts as long as the secrecy does, which is how a beverage formula has stayed proprietary for more than a century. The trade-off against patents is clean. A patent requires you to publish the invention in exchange for a time-limited right to exclude. A trade secret keeps the information but gives you nothing against a competitor who independently invents the same thing or lawfully reverse engineers your product.

Key idea: Patents cover inventions and turn on filing dates and disclosure; trademarks cover names and turn on distinctiveness and clearance; copyright is automatic but needs written assignments from contractors; trade secrets last forever and stop nobody who invents it independently.

Two agreements to handle early

Have employees and contractors sign a proprietary information and inventions agreement assigning work-related intellectual property to the company. And check the agreement you signed with your own current or former employer, because employment contracts frequently assign inventions created during employment, sometimes broadly. Founders occasionally discover after launch that a previous employer has a claim on the thing they built at night. That is a question to resolve with an attorney before you raise money or sign a customer contract, not during diligence.

Try it

Two friends are launching a specialty sauce company. They have a recipe, a distinctive brand name, a logo drawn by a freelancer, and plans to sell at farmers markets and eventually to grocery chains. What should they do, and where do they need a professional?

Answer: Entity: with two owners, food liability, and no plan to raise venture capital, a multi-member LLC with a written operating agreement is the usual starting point, and whether an S election helps depends on profit levels, which is a CPA question. Trademark: the brand name is the asset that will carry the business, so run a clearance search and consider a federal application early, before labels and signage are printed. Copyright: get a signed assignment from the freelancer for the logo immediately, because absent one the designer likely owns it. Recipe: this is a trade secret rather than a patent candidate in almost all cases, so protect it through limited access and confidentiality agreements with anyone who handles it, including co-packers. Regulatory: food production is licensed and inspected, and the rules differ by state and by whether they use a commercial kitchen or a co-packer. Professional required for: the operating agreement, the trademark filing, the co-packer contract, and the food licensing path.

Common misconceptions

• "An LLC means I cannot be sued personally." It does not cover your own conduct, guaranteed debts, or withheld payroll taxes, and it can be pierced.
• "An S corporation is a type of company." It is a federal tax election with eligibility limits, made by an existing corporation or LLC.
• "I paid for the logo, so I own it." A contractor generally retains copyright without a written assignment.
• "I can patent it after launching." In the United States a one-year grace period applies, but most other countries apply absolute novelty and foreign rights are lost on public disclosure.
• "A patent will protect my small business." A patent is a right to sue at your own expense, and enforcement costs are beyond most new ventures.
• "An EIN costs money." It is free directly from the IRS; paid services simply resell the same form.

Recap

• Sole proprietorship is the default; LLC is the usual answer with liability exposure; the S election is a tax question; C corporations are required for institutional investment.
• Limited liability fails against your own acts, personal guarantees, withheld payroll taxes, and alter-ego veil piercing.
• Separate bank accounts, real records, an operating agreement, and insurance are the practical protections.
• Utility patents last twenty years from filing, turn on filing dates, and public disclosure destroys most foreign rights.
• Trademark strength runs from fanciful down to generic, and clearance searches belong before naming rather than after.
• Copyright is automatic but contractors keep it without a written assignment, and trade secrets last indefinitely while giving no protection against independent invention.

Sources

1. U.S. Small Business Administration. (2025). Choose a business structure. sba.gov
2. Internal Revenue Service. (2025). Business structures. irs.gov
3. United States Patent and Trademark Office. (2025). Patent basics. uspto.gov
4. United States Patent and Trademark Office. (2025). Trademark basics. uspto.gov
5. U.S. Copyright Office. (2025). What is copyright? copyright.gov

The big picture

Ask founders who have failed what killed the company and a surprising number will not say the market, the product, or the money. They will say the other person. Noam Wasserman, who studied roughly ten thousand founders for his research on founding teams, argued that a large share of preventable startup failures trace to people decisions rather than to market conditions, and the decisions are made early, quickly, and badly.

His most quoted finding is almost comic in its ordinariness. About three quarters of founding teams split the equity within the first month, frequently in a single conversation, often equally, and usually before anyone knows who will actually do what, who will still be there in a year, or whether the business will be a software company or a consultancy. The split is then treated as permanent, because renegotiating it feels like an accusation. A decision made in week three governs the rest of the company's life.

This lesson covers the machinery that prevents that: whether to have a cofounder at all, how to split equity in a way you can defend, and the two mechanisms, vesting and written agreements, that let an early decision survive contact with reality.

Do you need a cofounder?

The advice that you must have one is stated far more confidently than the evidence supports. It is partly an investor preference, because a team is more robust to one person quitting and provides someone to talk to during diligence, and partly a genuine observation that the workload is brutal alone. Solo founders build successful companies routinely, including large ones.

The real question is what a cofounder buys you and what it costs. It buys complementary skills, someone who will argue with you before customers do, and resilience when one of you is having a terrible quarter. It costs a large fraction of the company, slower decisions, and a relationship that is unusually hard to exit; you cannot easily buy out a cofounder who holds a third of the business and has stopped contributing.

If you take one, work together on something real first. Not a hackathon weekend: a project with a deadline, a disappointment, and a disagreement in it. You are looking for how the person behaves when tired and wrong, which is a fact you cannot get from enthusiasm over coffee. Then have five conversations explicitly, out loud, before any equity is issued:

• Ambition. Are we building a company that supports two families comfortably, or one we intend to sell for a great deal of money? These are different businesses and both are respectable.
• Time. Who is full-time, from when, and what does part-time actually mean in hours?
• Money. How long can each of us go without salary? Who needs to draw first, and how much?
• Decisions. Who decides when we disagree, and about what? A fifty-fifty company with no tiebreak has a permanent deadlock risk.
• Departure. What happens to equity, customers, and the company if one of us leaves in month eight?

Key idea: A cofounder is a large permanent transaction, not an accessory. Work together on something real first, and settle ambition, time, money, decision rights, and departure before any shares are issued.

Splitting the equity

Roughly half of founding teams split equally, and equal splits have genuine virtues: they are simple, they signal mutual respect, and they avoid a corrosive negotiation at the start. They are also frequently wrong, specifically when contributions are highly asymmetric: one person full-time and one on evenings, one contributing the customer relationships and the capital, one joining eight months after the work began.

A more defensible approach is to make the factors explicit. Assign points, argue about the weights openly, and let the number fall out. Here is a worked version for two founders:

Factor	Weight	Founder A	Founder B
Full-time commitment from launch	40 points	40	40
Capital contributed (5 points per 10,000 dollars)	up to 20	10 (20,000 dollars)	0
Idea and prior work brought in	10 points	10	0
Domain expertise and existing customer relationships	10 points	10	0
Critical technical capability to build the product	10 points	0	10
Total		70	50

Seventy of one hundred twenty is 58.3 percent, and fifty is 41.7 percent, which most teams would round to sixty and forty. Now the honest caveat, because this table can mislead if you take it too seriously: the weights are invented. There is no correct number of points for an idea. The exercise is valuable because it forces the conversation into the open and produces a split each person can explain, not because the arithmetic is objective.

Two systematic biases to watch. First, idea origination is overvalued by the person who had the idea, essentially always. Ideas are abundant and the plan will change; the person who works full-time for three years is contributing far more. Weight commitment heavily and origination lightly. Second, past contribution is easy to measure and future contribution is what matters, so a split that rewards the last six months while ignoring the next four years will be wrong within a year. Dynamic models such as continuously recalculated splits attempt to fix this; they exist, they solve a real problem, and they introduce ongoing accounting that becomes its own source of argument. Vesting solves most of the same problem with far less friction.

Key idea: Make split factors explicit and weight full-time commitment heavily and idea origination lightly. The value of the exercise is the conversation, not the precision of the number.

Vesting, the mechanism that saves companies

Vesting means you earn your shares over time rather than owning them outright at issue. The standard structure is four years with a one-year cliff: nothing vests for the first twelve months, then twenty-five percent vests at once on the anniversary, then the remainder vests monthly over the following thirty-six months.

The essential point, and the one first-time founders resist, is that founders should vest too. Consider what happens without it. Two cofounders split fifty-fifty, one leaves after five months because the work is harder than expected, and they keep half the company forever. The remaining founder now does all the work for half the outcome, and every future investor sees a large permanently inactive block on the cap table, which they will insist be fixed before they invest. It cannot be fixed without the departed founder's consent, and their incentive to consent is zero. Vesting converts that catastrophe into a footnote: five months of a four-year schedule with a twelve-month cliff means they leave with nothing, which is the correct answer.

Acceleration covers what happens if the company is acquired. Single-trigger acceleration vests unvested shares on the acquisition itself. Double-trigger vests them only if the acquisition happens and the person is terminated afterward. Double-trigger is the norm for founders, because buyers do not like paying for a team that vests fully on closing and then leaves.

Now the deadline that costs people real money. When you receive restricted stock subject to vesting, United States tax law generally treats each tranche as income as it vests, valued at that time. An 83(b) election lets you instead be taxed on the value at grant, which for a newly formed company is usually near zero. If the company later becomes valuable, the difference can be enormous. The election must be filed within thirty days of the grant, the deadline cannot be extended, and there is no fix afterward. This is exactly the kind of item that belongs on a calendar and in a conversation with a CPA or attorney at incorporation, not something to read about later.

Key idea: Four-year vesting with a one-year cliff, applied to founders as well as employees, is what prevents an early departure from permanently damaging the company. The 83(b) election has a hard thirty-day deadline.

Put it in writing

A founders' agreement should state roles and responsibilities, the equity split and vesting terms, who decides what and how deadlocks break, an assignment of all relevant intellectual property to the company, what happens on voluntary departure and on removal, and confidentiality. Write it while everyone is optimistic, because that is the only time it can be written fairly. Once there is something to fight over, every clause reads as an accusation.

First hires

The first classification question comes before the first hire. The IRS applies a common-law test with three families of evidence: behavioral control, meaning who directs how the work is done; financial control, meaning who supplies tools, bears expense, and can realize a profit or loss; and the type of relationship, meaning contracts, benefits, permanence, and whether the work is central to the business. Calling someone a contractor does not make them one, and misclassification exposes the business to back taxes, interest, and penalties. Several states apply stricter tests than the federal one. The IRS will make a determination on request, and this is a question to put to a professional before the first payment, not after a state audit letter.

Two practical notes on cost and choice. Budget an employee at roughly one and a quarter to one and a third times salary once payroll taxes, insurance, benefits, and equipment are counted. And hire against your binding constraint rather than against your preferences: the temptation is to hire away the work you dislike, but the correct first hire is whoever removes the ceiling on the business, which in the grooming van example is a groomer and in a sales-limited business is a salesperson.

On selection, the research is unusually clear and unusually ignored. Work sample tests, in which candidates do a small piece of the actual job, and structured interviews, in which every candidate gets the same questions scored against defined criteria, predict performance substantially better than the unstructured conversation most founders default to. The unstructured interview is among the weakest common predictors and among the most confidently trusted, which is a bad combination. Ask the same questions in the same order, score them, and give a paid short work sample where the job allows it.

On equity for employees, be honest rather than promotional. Options carry a strike price set by a valuation, they are usually worthless because most companies do not produce liquidity, the standard window to exercise after leaving has historically been ninety days, and the preference stack from the funding lesson sits ahead of common stock. A founder who explains all of that plainly and then offers options is treating an employee as an adult. A founder who describes options as if they were a bonus is setting up a grievance for later.

Key idea: Classification is determined by the facts of control, not by the label; hire against the binding constraint; use work samples and structured interviews; and describe equity honestly, including its usual outcome.

Try it

Two people start a company. A has worked in the industry nine years, brings the initial three customers, contributes 30,000 dollars, and goes full-time immediately. B is a strong engineer who will build the product, contributes no capital, and stays at their job for the first six months before going full-time. They are about to split fifty-fifty with no vesting. What would you advise?

Answer: The fifty-fifty split is defensible on future contribution, since B's engineering is essential and both will be full-time within six months, and an argument for something like sixty-forty is also defensible given A's capital, customers, and earlier full-time start. Either can work. What cannot work is the absence of vesting: if B decides after four months that they prefer their job, the company loses its engineer and permanently gives away half of itself. Advise four-year vesting with a one-year cliff for both, with B's clock starting when they go full-time or with a documented partial credit for the part-time period. Also advise: capital contributed as a documented loan rather than folded into the split, so it can be repaid rather than argued about; a written founders' agreement with a deadlock mechanism; intellectual property assigned to the company; and an 83(b) election filed within thirty days of the stock grant, confirmed with a CPA.

Common misconceptions

• "You need a cofounder." It is a strong investor preference and a genuine workload argument, not an evidence-backed requirement. Solo founders succeed regularly.
• "An equal split is always fairest." It is often right and often wrong, and the fairness comes from the conversation rather than the symmetry.
• "The idea deserves most of the equity." Ideas are abundant and the plan will change; sustained full-time work is the scarce contribution.
• "Vesting is something investors impose on employees." Founder vesting is what protects the remaining founder when a cofounder leaves early.
• "I can handle the 83(b) later." The window is thirty days from grant and cannot be extended.
• "Calling someone a contractor makes them one." Classification turns on behavioral control, financial control, and the relationship, and misclassification carries real penalties.

Recap

• Most founding teams split equity within a month, often before anyone knows who will do what, and rarely revisit it.
• Settle ambition, time, money, decision rights, and departure terms before issuing shares, ideally after working together on something real.
• Explicit split factors produce a defensible number; weight full-time commitment heavily and idea origination lightly.
• Four-year vesting with a one-year cliff, applied to founders too, prevents an early departure from permanently damaging the company.
• Double-trigger acceleration is the founder norm, and the 83(b) election must be filed within thirty days of grant.
• Worker classification follows the facts of control; hire against the binding constraint; work samples and structured interviews beat unstructured conversation.

Sources

1. Internal Revenue Service. (2025). Independent contractor or employee. irs.gov
2. U.S. Small Business Administration. (2025). Hire and manage employees. sba.gov
3. Wasserman, N. (2008). The founder's dilemma. Harvard Business Review, 86(2), 102-109. hbr.org
4. U.S. Department of Labor. (2025). Employer responsibilities and wage and hour guidance. dol.gov
5. Wikipedia contributors. (2025). Vesting. en.wikipedia.org

The big picture

A large number of technically capable founders quietly believe selling is somebody else's job, or worse, a slightly disreputable activity they will outsource once the product is good enough. This belief kills more good products than competition does. It also misunderstands what early selling is. Talking to prospects with a price attached is customer discovery that has become expensive to ignore: the objection you hear on the twelfth call is the product change you needed to make, and only the founder can act on it. Nobody will sell your first hundred customers better than you, not because you are charismatic, but because you are the only person who can rewrite the product in response to what you hear.

This lesson is about doing that systematically. Most of it is arithmetic, which is good news if you are one of the people who dreads selling: a pipeline is a calculation before it is a personality.

Channel economics decides everything

The single most common go-to-market error is choosing a channel that costs more than the product can support. The rule underneath it is simple: the cost of the sales process must fit the price and margin of the thing being sold.

Work it. A salesperson fully loaded costs perhaps 110,000 dollars a year. For that role to be worth having, the gross profit from their sales should comfortably exceed their cost, and a common planning target is roughly three times. Call it 330,000 dollars of gross profit.

• Selling a product at 6,000 dollars a year with an 80 percent gross margin produces 4,800 dollars of gross profit per customer. The rep needs about 69 customers a year, or under six a month. That is an ordinary quota and the model works.
• Selling a product at 360 dollars a year with the same margin produces 288 dollars per customer. The rep now needs about 1,146 customers a year, which is roughly 95 a month, or five every working day, forever. No human does this. The channel is impossible, and no amount of sales training fixes it.

So low-priced products require self-serve channels: the customer finds you, evaluates alone, and buys without talking to anyone. High-priced products can support human selling, and very high-priced products can support travel, long cycles, and multi-person buying committees. The channel is not a preference; it is implied by your price.

Annual price per customer	Channels that can work	Implied acquisition cost ceiling
Under 200 dollars	Self-serve, search, content, word of mouth, marketplaces	Tens of dollars
1,000 to 10,000 dollars	Inside sales by phone and video, targeted marketing, partnerships	Hundreds to low thousands
50,000 dollars and above	Field sales, long cycles, conferences, references	Tens of thousands

Key idea: Price determines the channel. A sales process costing thousands of dollars cannot sell a product worth hundreds, and the arithmetic is fixed before any selling begins.

Pipeline arithmetic turns a goal into a calendar

"We need customers" is not a plan. Work backwards instead.

Suppose you need 10 new customers this quarter. Your close rate on qualified opportunities is 20 percent, so you need 50 qualified opportunities. About one in four real conversations turns into a qualified opportunity, so you need 200 conversations. About one in five outreach attempts produces a conversation, so you need 1,000 attempts. Spread across roughly 65 working days in a quarter, that is about 16 outreach attempts a day.

Sixteen a day is a task. It goes on a calendar, it gets done or it does not, and at the end of the first month you know something either way. If the ratios are worse than assumed, you find out in week two rather than in month four, and you can change the message, the segment, or the channel while it is still cheap. The arithmetic also disciplines optimism: a founder who believes they will sign 10 customers next quarter while making four calls a week is not being ambitious, they are making an error they could have caught with a division.

Key idea: Pipeline math converts a customer target into a daily activity number through close rate, qualification rate, and response rate, and it exposes an impossible plan within two weeks.

Marketing with no money

Early marketing options are more limited and more effective than founders expect. Ordered roughly by how well they serve a new business with no budget:

• Referrals, systematized. Not "we hope for word of mouth" but a specific habit: ask every satisfied customer, at the moment of satisfaction, for one introduction by name. This is the cheapest channel that exists and almost nobody works it deliberately.
• Partnerships with adjacent businesses. The veterinary clinic that hands out the grooming van's cards, the accountant who refers bookkeeping clients, the wedding venue that recommends a florist. Your customers are already someone else's customers.
• Being where your customers already gather. Trade associations, professional forums, local groups, industry conferences. Presence over promotion: answer questions for six months and you become the obvious person to call.
• Content and search. Slow and compounding. It works when people search for the problem you solve, and it does nothing for a problem nobody knows they have.
• An email list. The only audience you own outright, unaffected by another company's algorithm changes.
• Local and physical. For local services, signage, sponsorships, and direct mail still work, and are frequently ignored by founders who assume everything must be digital.

Two honest notes. Paid advertising is usually a poor first channel: your acquisition cost is unknown, you are bidding against companies who know theirs, and a thin-margin business cannot absorb the learning period. It becomes an excellent channel later, once you know your contribution margin and payback period, because at that point it is a dial rather than a gamble. And press is systematically overvalued: a launch story produces a spike, a pleasant afternoon, and almost no durable customer flow. It supports credibility; it is not a channel.

One legal point worth stating plainly: advertising claims must be truthful and substantiated. The Federal Trade Commission requires that objective claims about performance, results, or comparisons have evidence behind them before they are made, and rules about endorsements and testimonials apply to small businesses too. Making up a statistic for a landing page is not a growth hack.

Doing things that do not scale

Paul Graham's well-known argument is that early companies should recruit users manually, one at a time, in ways that could never work at scale. The examples are famous. The Airbnb founders travelled to New York to meet hosts in person and photograph their apartments themselves, because the listings had bad photographs and no algorithm was going to fix that. The founders of Stripe would offer, when someone expressed interest, to set the product up on the spot rather than emailing a signup link, removing the last step of friction by doing it themselves.

The tactic is genuinely good practice, for a reason that has nothing to do with heroism: manual recruitment puts you in the room. You watch the customer fail to understand the thing you thought was obvious, and you learn what to change. Automation cannot give you that.

Now the caveat this course owes you. These are survivor stories, told by the winners, and they get retold as though the tactic caused the outcome. It did not. A great many founders knocked on doors, personally installed software, and hand-delivered products to a market that did not care, and no one wrote about them. The honest version is narrower and still worth having: manual, unscalable recruitment is the fastest way to learn what your product should be, and it improves your odds. It does not confer the outcome, and the Airbnb story in particular is usually told without the part where the company nearly died repeatedly and the founders funded themselves selling novelty cereal.

Key idea: Recruit your first customers by hand because it teaches you what to change, not because the famous cases prove it works. Those are survivor accounts, and the same tactics failed quietly for many others.

The mechanics of an early sale

Five habits carry most of the value.

Qualify before you pitch. Establish whether this person has the problem, cares about it, can authorize spending, and has any reason to act this quarter. Crude checklists like budget, authority, need, and timing are heuristics rather than laws, and their real function is to stop you spending three months on someone who was never going to buy.

Diagnose before you prescribe. The discovery questions from Module 2 belong here too. A prospect who has just described their own problem out loud is far easier to sell to than one who has been talked at.

Answer objections with questions. "It is too expensive" can mean it costs more than the budget, more than the perceived value, or more than the alternative. Those need three different responses, and arguing before you know which one it is loses the deal.

Ask for the order. Explicitly, with a specific next step and a date. An enormous share of early sales are lost because the founder finished the demo, said they would follow up, and never proposed anything.

Follow up more than feels comfortable. Most deals close after several contacts, and most founders stop after one because a second feels like pestering. It usually is not; the prospect is busy and your email is not the most important thing in their week.

One pricing discipline: resist discounting to win early logos. A discount sets a reference price you will fight for years, and word travels between customers. If you need to reduce risk for an early buyer, use a shorter term, a paid pilot, or a narrower scope, all of which preserve the price.

What the first hundred customers are for

They are not primarily a revenue source. Handpick them, over-serve them beyond what is economical, and interview them constantly. Their job is to teach you what to build, what to say, and who else looks like them. The economics of serving them badly do not matter yet; the economics of never learning what they need are fatal.

And keep the churn arithmetic from the projections lesson in view. Since the steady state equals additions divided by churn, keeping a customer moves the ceiling more than winning one. The most underrated growth activity available to a young company is calling the customers who left and finding out why.

Key idea: Handpick and over-serve the first hundred customers to learn from them, and treat retention as a growth channel because it moves the ceiling more than acquisition does.

Try it

You sell a 4,800 dollar annual service. You need 24 customers in the next year. Your close rate on qualified opportunities is 25 percent, one in three conversations becomes qualified, and one in six outreach attempts produces a conversation. How much activity per working day does the plan require, and what does that tell you?

Answer: Twenty-four customers at a 25 percent close rate requires 96 qualified opportunities. At one in three, that is 288 conversations. At one in six, that is 1,728 outreach attempts, which over roughly 250 working days is about 7 per day. That is a sustainable individual workload, so the plan is feasible for one founder selling part-time alongside delivery, which is the real check. If the same calculation had produced 40 attempts a day, the answer would be that this target requires either a dedicated salesperson, a higher-converting channel, or a smaller target, and knowing that in January is worth a great deal more than discovering it in September.

Common misconceptions

• "A good product sells itself." Distribution is a separate problem from quality, and the founder is the only person who can change the product in response to objections.
• "We will hire a salesperson to fix growth." A salesperson cannot rescue a channel the price cannot support, and cannot make the product changes early objections call for.
• "Advertising is how you get customers." It is a dial you turn once you know your unit economics, and a poor first channel while acquisition cost is still unknown.
• "Press coverage will launch us." It produces a spike and credibility, not a durable channel.
• "Following up repeatedly is rude." Most deals close after several contacts, and most early sales are lost to silence rather than to rejection.
• "The famous manual-growth stories prove the tactic works." They are survivor accounts; the tactic teaches you a great deal and guarantees nothing.

Recap

• Channel economics is decided by price: a 110,000 dollar salesperson works at a 6,000 dollar price point and cannot work at 360 dollars.
• Pipeline arithmetic converts a customer target into a daily activity number and exposes impossible plans within weeks.
• Referrals, adjacent partnerships, presence in existing communities, content, and an owned email list are the workable no-budget channels.
• Paid advertising belongs after unit economics are known; press produces a spike rather than a channel; advertising claims must be substantiated.
• Manual customer recruitment is valuable because it teaches you what to change, and the famous examples are survivor accounts rather than proof.
• Qualify, diagnose, answer objections with questions, ask for the order, follow up, and protect the price with pilots rather than discounts.

Sources

1. U.S. Small Business Administration. (2025). Marketing and sales. sba.gov
2. Graham, P. (2013). Do things that don't scale. paulgraham.com
3. Federal Trade Commission. (2025). Advertising and marketing guidance. ftc.gov
4. Shepherd, D. A., et al. (2020). Marketing and sales for the entrepreneur. In Entrepreneurship. OpenStax, Rice University. openstax.org
5. SCORE Association. (2025). Sales and marketing resources for small business. score.org

The big picture

Growth is treated as the reward at the end of the hard part. It is closer to the opposite: a new and less familiar set of problems that arrive precisely when everyone is celebrating. The bakery that finally sells 120 loaves a day discovers it needs a second oven, a second baker, and a bigger flour order, all payable before the extra bread is sold. The cleaning company that lands four new contracts discovers it must hire and pay crews for six weeks before the first invoice clears. Nothing has gone wrong. This is what success does to a bank account.

This lesson covers the second half of a venture's life: the mechanics of growth, the honest evidence about why companies die, and how to tell a disciplined change of direction from panic wearing a business word.

Growth eats cash, worked

Take a commercial cleaning company. Revenue is 50,000 dollars a month and about to grow 20 percent a month. Variable costs are 65 percent of revenue, mostly crew wages, and they are paid in the month the work is done. Fixed costs are 8,000 dollars a month. Customers are invoiced at month end on 45 day terms, so cash arrives roughly two months after the work.

Month	Revenue	Cash out	Cash in	Net cash	Accounting profit
1	50,000	40,500	50,000	+9,500	+9,500
2	60,000	47,000	50,000	+3,000	+13,000
3	72,000	54,800	50,000	-4,800	+17,200
4	86,400	64,160	60,000	-4,160	+22,240
5	103,680	75,392	72,000	-3,392	+28,288
6	124,416	88,870	86,400	-2,470	+35,546

Look at the last two columns together. In month six this company earned 35,546 dollars of profit and its bank balance fell by 2,470 dollars. It has lost cash in four consecutive months while growing 20 percent a month at a 35 percent gross margin. Nothing here is a mistake; the cash simply arrives two months after the wages that produced it, and each month's wages are larger than the last.

The levers are all about timing rather than about margin. Invoice on the service date rather than at month end. Shorten terms to net fifteen, or take a deposit, or offer a two percent discount for payment within ten days. Ask suppliers for longer terms so your own outflows move later. Arrange a line of credit before you need one, because lenders are far more willing when you do not. Or, and this is the option founders resist most, grow more slowly: 10 percent a month with the same terms would have stayed cash positive throughout.

Key idea: Growth consumes working capital because costs are paid before revenue is collected. A profitable, fast-growing business can lose cash every month, and the fixes are all about timing, not margin.

Premature scaling and the founder's own job

Premature scaling is spending like a proven business before the model is proven: hiring ahead of revenue, opening a second location before the first throws off reliable cash, expanding to a new city before the first is profitable, building a brand campaign before anyone renews. An industry study by the Startup Genome project argued that a large majority of failed high-growth internet startups showed this pattern. That report was not peer reviewed and its sample was self-selected, so treat the specific number as directional. The mechanism is easy to verify without it: every one of those moves converts flexible cost into fixed cost, and fixed cost is what removes your ability to survive a bad quarter.

The other thing that changes with size is your own job. What made you effective at three people, doing everything yourself and knowing every detail, is exactly what breaks at fifteen. The transition requires delegating work you are still the best at, hiring people to manage other people, and writing down processes that lived in your head. Larry Greiner's 1972 model of organizational growth describes this as a sequence of phases each ending in its own crisis: a crisis of leadership when founders can no longer coordinate personally, then of autonomy, then of control, then of red tape. Take the model loosely. It is a descriptive vocabulary from half a century ago, not a predictive schedule, and companies skip stages and go backwards. Its real value is the reminder that the problems do not just get bigger, they change shape, and the founder who refuses to change with them becomes the constraint.

Key idea: Premature scaling converts flexible costs into fixed ones before the model is proven, and the founder's job changes qualitatively with size rather than just getting larger.

What actually kills companies

The most cited evidence here is CB Insights' analysis of startup failure post-mortems, in which founders wrote publicly about why their companies died. The frequently reported causes include running out of cash or failing to raise more, cited in roughly 38 percent of cases; no market need, roughly 35 percent; getting outcompeted, roughly 20 percent; a flawed business model, roughly 19 percent; regulatory or legal problems, roughly 18 percent; pricing or cost issues, roughly 15 percent; not having the right team, roughly 14 percent; poor timing, roughly 10 percent; a poor product, roughly 8 percent; and disharmony among the team or with investors, roughly 7 percent. The figures sum well past 100 because most failures had several causes.

Now cite it carefully, because this list is quoted everywhere as if it were a base rate and it is not. The sample is a small number of mostly venture-backed technology companies whose founders chose to publish an account of failing. That selection is not random in any direction you can correct for: founders who write post-mortems are unusually reflective, unusually public, and unusually likely to be in a community where that is expected. What the list gives you is a catalogue of plausible mechanisms, and it is genuinely useful as that. What it does not give you is the probability that any of these will kill your bakery.

For the ordinary business, the evidence points at one thing above all: cash. Analysis by the JPMorgan Chase Institute of millions of small business bank accounts found that the median small business holds a cash buffer of roughly 27 days, meaning less than a month of outflows in reserve. Sit with that number. A slow season, one anchor client leaving, a customer paying sixty days late, or a compressor failing is enough to end a business that was otherwise fine. It also explains why the cash-flow gap lesson, which looks like bookkeeping, is the most consequential material in this course.

Two other killers deserve naming. Cofounder conflict destroys companies that had working products, which is why the vesting and written-agreement material in Module 5 matters so much. And key-person dependence, where every important customer relationship and every undocumented process lives in one person's head, turns any illness, burnout, or departure into an existential event.

The honest note about "no market need": it is the second most cited cause on that list, and preventing it is the entire purpose of Modules 2 and 3. It still happens constantly, to careful people, because customer discovery reduces the risk rather than eliminating it. Do the work anyway; a reduced risk on a bet this large is worth a great deal.

Key idea: The famous post-mortem list is a catalogue of mechanisms from a self-selected venture sample, not a base rate. For ordinary businesses the dominant killer is cash, with a median buffer of under a month.

Pivots done well and badly

A pivot is a structured change to one core hypothesis while keeping what you have already validated. That definition does real work, because most things called pivots are not that. If you change the customer, the problem, the product, and the model at once, you have not pivoted; you have started a different company with the same bank account.

The famous examples are worth knowing and worth deflating slightly. Slack emerged from Glitch, a game company that failed; the internal chat tool the team had built for themselves became the product. Instagram began as Burbn, a cluttered check-in application, and became a company when the founders deleted everything except photos, filters, and comments. Shopify began as an online snowboard shop whose founders built their own store software and realized the software was the business. Twitter emerged from Odeo, a podcasting company whose market was undercut when Apple built podcast support into iTunes.

Notice what these accounts usually omit. Slack's founder had done this before, having built Flickr out of an earlier failed game, and he had unusual credibility with investors and enough remaining capital to try again; he had also offered to return money to investors rather than spend it on a hope. Odeo's investors were offered the chance to buy back their stake. In every case the pivot succeeded partly because the team had runway, reputation, and a specific new hypothesis in hand. The many companies that pivoted from a failing product into a second failing product did not get written about.

Here is a usable test.

Disciplined pivot	Thrashing
One hypothesis changes; validated learning is retained	Everything changes at once
Triggered by a failed pre-registered experiment	Triggered by exhaustion or a conference talk
New hypothesis is specific and testable	New direction is a category, not a hypothesis
Enough runway remains to test it properly	Runway is two months and the test needs six
The team is told the reasoning and the evidence	The team learns about it from an announcement
First one in eighteen months	Third one this year

And know when to persevere, because premature pivoting is as common as stubbornness. Stay the course when cohort retention is improving even if totals are flat, when the problem is execution rather than demand, when customers complain loudly if the product breaks, which is the single most encouraging signal a young company can get, and when your experiments are still returning new information. Change course when repeated well-designed experiments come back negative, when the customers you do have never renew, or when the arithmetic itself does not close no matter how well you execute.

Key idea: A pivot changes one hypothesis and keeps the learning. Disciplined pivots follow failed pre-registered tests and have the runway to test the replacement; thrashing changes everything, repeatedly, on no new evidence.

Try it

A two-year-old software company has 60 customers paying 200 dollars a month, monthly churn of 9 percent, four months of runway, and a founder who has just returned from a conference convinced the company should sell to hospitals instead. Diagnose.

Answer: First, the arithmetic. Nine percent monthly churn against whatever new sales they add gives a steady state of additions divided by 0.09, and at 60 customers the base is losing five or six a month, so the current business is not compounding. The problem is retention, not market choice. Second, the proposed pivot fails every test in the table: it changes customer, product requirements, sales cycle, and regulatory exposure simultaneously; it was triggered by a conference rather than an experiment; and hospital sales cycles routinely exceed twelve months against four months of runway, so the new hypothesis cannot be tested before the money runs out. The correct next action is to spend two weeks interviewing the customers who churned, because a 9 percent monthly churn rate is a specific, diagnosable problem, and to build the funding or cost plan that buys enough runway to act on what they learn. If retention cannot be fixed and the segment genuinely does not need the product, then a pivot is warranted, chosen on that evidence and scoped to something testable in the runway available.

Common misconceptions

• "Profitable businesses do not run out of cash." Growth pays costs before it collects revenue, and four consecutive profitable months can drain the account.
• "Faster growth is always better." Growth converts flexible costs to fixed ones and consumes working capital; the right rate is the one your cash can fund.
• "The startup failure percentages are base rates." They come from a self-selected sample of mostly venture-backed founders who chose to write about failing.
• "Companies die of bad products." Poor product is among the least cited causes; cash and absent demand dominate.
• "Pivoting is what smart founders do." The famous pivots had runway, reputation, and a specific new hypothesis; most pivots are a second failing product.
• "Flat total users means it is time to change direction." Improving cohort retention under flat totals is a reason to persevere, not to pivot.

Recap

• The cleaning company earned 35,546 dollars of profit in month six while its cash fell, because collections lag the wages that produced them.
• The fixes for a growth cash gap are timing levers: invoice sooner, shorten terms, take deposits, lengthen payables, arrange credit early, or grow more slowly.
• Premature scaling turns flexible costs into fixed ones before the model is proven, and the founder's job changes qualitatively as the company grows.
• Post-mortem data lists cash, no market need, competition, and flawed models as leading mechanisms, from a self-selected venture-heavy sample.
• The median small business holds roughly 27 days of cash buffer, which is why cash management is the decisive small business skill.
• A disciplined pivot changes one hypothesis, follows a failed test, is specific, and has runway; thrashing changes everything repeatedly on no new evidence.

Sources

1. CB Insights. (2021). The top reasons startups fail: analysis of startup post-mortems. cbinsights.com
2. JPMorgan Chase Institute. (2016). Cash is king: Flows, balances, and buffer days. jpmorganchase.com
3. Wikipedia contributors. (2025). Lean startup. en.wikipedia.org
4. U.S. Small Business Administration. (2025). Grow your business. sba.gov
5. SCORE Association. (2025). Cash flow management resources. score.org

The big picture

Every business ends. That sentence sounds grim and is not: it simply means that the question of how a venture concludes deserves as much thought as how it starts, and almost never gets any. Founders spend months on a launch plan and no time at all on the far more likely scenarios, which are that the business runs profitably for twenty years, or is sold quietly to a competitor for a modest sum, or is wound down deliberately when the owner is ready to do something else.

This closing lesson covers the endings, two paths into ownership that the entrepreneurship literature underweights, and then the only question that actually matters to you personally: given everything in this course, should you do this?

How businesses actually end

The endings, roughly by frequency:

• It keeps operating. The most common outcome for a successful small business is that it continues, pays its owner, and is never sold. This is an ending nobody calls an exit and it is the modal good result.
• A small sale. To a competitor, a supplier, a customer, an individual buyer, or a private equity firm rolling up an industry. Most acquisitions are unremarkable transactions in the hundreds of thousands to low millions of dollars.
• Sale to employees, management, or family. A management buyout, an employee stock ownership plan with its particular tax treatment, or succession within a family, which fails more often than expected because the successor's willingness was assumed rather than established.
• A deliberate wind-down. Pay the debts, tell customers early, help the staff find work, and file the dissolution. A respectable ending, treated as shameful far more often than it deserves.
• An initial public offering. The one everybody pictures. A few hundred United States listings a year at most, and far fewer traditional operating-company offerings, against roughly five million business applications. It is a rounding error as an outcome.

Key idea: Continuing to operate is the most common good ending, a modest sale is the most common transaction, and a public offering is statistically negligible.

What a small business is worth

Small business valuation usually runs on seller's discretionary earnings: net profit, plus the owner's salary, plus owner benefits and personal expenses run through the business, plus interest, depreciation, amortization, and genuinely one-time costs. The idea is to show a buyer what the business would produce for a new owner-operator.

Main street businesses commonly transact at roughly two to three times SDE. Larger lower-middle-market companies are priced on EBITDA instead, commonly in the three to six times range, with higher multiples for scale, recurring revenue, and diversified customers. Take the plumbing firm from Lesson 1: 2.4 million dollars of revenue and about 290,000 dollars of SDE. At 2.5 times, that is roughly 725,000 dollars.

Here is the part worth internalizing years before you sell. The multiple is not fixed; it is a judgment about risk, and you control most of the inputs.

Raises the multiple	Lowers the multiple
Recurring contracts and predictable revenue	One-off project work
Many customers, none dominant	One customer at 40 percent of revenue
A manager who runs operations	Revenue that depends on the owner personally
Documented processes and clean books	Knowledge in one person's head, informal records
Trained staff who intend to stay	Key employees likely to leave with the seller
Growth with evidence behind it	Deferred maintenance and aging equipment

Read that table as an instruction rather than a description. The work of making yourself replaceable, documenting how things are done, hiring a manager, spreading customer concentration, is exactly the same work that turns a job you own into an asset you own. It also makes the business better to run in the meantime.

On structure, three honest notes. Seller financing, where the seller takes part of the price as a note paid over years, is common and means you may not be fully paid at closing. Earnouts tie part of the price to future performance you may no longer control. And asset sales and stock sales carry large, asymmetric tax consequences for buyer and seller, which is squarely a matter for a CPA and an attorney.

Key idea: Small businesses commonly sell for two to three times seller's discretionary earnings, and the multiple rises with recurring revenue, customer diversity, documented processes, and independence from the owner.

Social ventures and nonprofits

Three things get conflated constantly, so separate them.

Nonprofit status is a tax classification, not a business model. A 501(c)(3) organization is subject to a nondistribution constraint: no individual may take the profits, which is the actual meaning of the term, rather than any requirement that the organization avoid earning money. Many nonprofits earn most of their revenue from services. The structure requires a board with real governance duties, an application to the IRS, and ongoing annual filings.

A benefit corporation is a for-profit legal entity form available in most states, which changes directors' duties so they must consider stakeholders and stated public benefits alongside shareholders. A certified B Corporation is something else entirely: a private certification awarded by a nonprofit organization after an assessment. A company can be one, the other, both, or neither, and the two are routinely reported as if they were the same thing.

Now the part that makes this a lesson rather than a glossary. Social ventures are unusually prone to measuring activities instead of outcomes, because the activities are visible and the outcomes are hard. The field's own best example is microcredit. Small loans to poor entrepreneurs were widely described as transformative, and Muhammad Yunus and Grameen Bank received the Nobel Peace Prize in 2006. Then researchers ran the experiment properly. Six randomized controlled trials across different countries, published together in 2015, found modest increases in business investment and borrowing, and little or no average effect on household income, consumption, health, or children's schooling.

The honest reading is neither triumph nor debunking. Microcredit turned out to be a useful financial product that gives poor households more control over timing and risk, and not a cure for poverty. What deserves genuine admiration is that the field tested its own flagship claim and published the disappointing result. That is the standard to hold yourself to: state your intended outcome in measurable terms, measure it against a comparison, and be willing to report that it did not work.

And one piece of arithmetic that social missions do not exempt you from. A venture that cannot cover its costs from revenue is a charity, which is an honourable and necessary thing to be, and it should then be funded as one, with a fundraising plan rather than a business plan. Confusing the two produces organizations that fail at both.

Key idea: Nonprofit is a tax status with a nondistribution constraint, benefit corporations and B Corp certification are different things, and social ventures must measure outcomes rather than activities and still close their own unit economics.

Buying a business instead of starting one

This is the most underrated path in entrepreneurship, and it follows directly from Lesson 2. Starting a business means accepting the base rate: roughly a third of new establishments reach year ten. Buying one means acquiring a business that has already survived, with customers, staff, systems, supplier relationships, and, most importantly, cash flow from day one. You skip the entire cash-flow gap.

The demographic conditions favour it. A large cohort of owners built businesses over decades and is now approaching retirement, and a substantial share have no succession plan and no family successor. Financing is available in ways it is not for startups, because the business has a history a lender can underwrite: SBA 7(a) loans are widely used for acquisitions, and seller financing frequently covers part of the price.

At the larger end sits the search fund model, in which investors back an individual to spend a year or two searching for a company to buy and then run. Studies from Stanford's business school have reported strong aggregate returns for this asset class, with pre-tax internal rates of return across the studied population in the mid-thirty percent range. Read those figures with the same care you apply to venture returns. The aggregate is driven by a minority of outcomes; a meaningful share of searchers never acquire anything at all; and among those who do acquire, a substantial minority lose money or return less than the capital invested. It is a power law wearing a quieter suit.

The risks are specific and diagnosable. You can overpay. Revenue may walk out the door with the seller who held every relationship. Customer concentration can hide in a summary income statement. Maintenance may have been deferred for three years to inflate earnings before a sale. The standard defences are a quality of earnings review by an accountant, a written transition period with the seller, and a price structure that keeps some of the money contingent.

Key idea: Buying an existing business skips the survival base rate and the cash-flow gap, is financeable in ways startups are not, and requires real diligence against overpaying, owner-dependent revenue, and deferred maintenance.

So, should you start?

Here is the framework, and it is deliberately unromantic.

Opportunity cost. What does your best alternative pay over the next three years? Recall from Lesson 2 that median self-employment earnings run below comparable wage employment, and that the payoff is concentrated in a right tail. You are not required to beat the median. You are required to know what you are giving up.

Downside tolerance. How many months can your household run without your income? What happens to health coverage? Do you have dependents, debt payments, or an immigration status tied to employment? These are not obstacles to be overcome by determination; they are constraints that determine which version of this you can responsibly attempt.

The guarantee question. Will you be asked to sign a personal guarantee, and do you understand what default would mean for your household? A founder who has not answered this has not priced the downside.

Reversibility. If this fails in eighteen months, can you get an equivalent job? For most people in most fields the answer is yes, and that fact is far more encouraging than any founder anecdote.

Motivation. Are you moving toward something specific or away from a job you dislike? Both are common. Only the first predicts persistence through the eleventh month, when nothing is working and no one is watching.

There are situations where the honest answer is not now: no savings buffer and no other income; a medical situation dependent on employer coverage; a partner who has not genuinely agreed; an idea you have not tested in a single customer conversation; or a plan that only works if a hockey stick materializes. None of these are permanent conditions, and treating them as reasons to wait is a strategic decision rather than a failure of nerve.

Which points to what most successful founders actually do, whatever the mythology says. Keep the job. Run the customer interviews from Module 2 on evenings and weekends. Run one pre-registered experiment. Take the first paying customer while employed, checking your employment agreement first, since as Lesson 12 noted it may assign inventions or restrict outside work. Leave when the revenue or the evidence justifies leaving, not when the enthusiasm peaks. This is the staged path, it is available to almost everyone, and it converts an irreversible bet into a series of small reversible ones.

And if the answer is not now, the waiting is not wasted. Go and get the domain knowledge, because Lesson 3 showed that is where opportunities come from. Build the savings buffer that determines your downside tolerance. Build the network that becomes your first channel. Learn to sell, which is the skill that transfers to everything.

Key idea: Decide on opportunity cost, downside tolerance, the guarantee, reversibility, and motivation, and prefer the staged path that converts one irreversible bet into a sequence of small reversible ones.

A last word on what this course was about

We began with base rates and survivorship bias because entrepreneurship is taught almost everywhere as a highlight reel, and a highlight reel is a bad map. Most new businesses do not survive ten years. Most venture-backed startups return nothing. The famous stories are drawn from a sample filtered by success, and their lessons are correspondingly unreliable.

None of that is an argument against doing it. It is an argument for doing it with arithmetic: a contribution margin you have calculated, a cash-flow gap you have sized, a customer you have actually spoken to, an experiment with a threshold you wrote down first, and an honest account of what you are risking. Those habits will not guarantee an outcome. They will move your position within the distribution, which is the only thing anyone can actually offer you.

And keep the ordinary business in view. The bakery, the plumbing firm with eight vans, the two-person consultancy, the woman who noticed her town had no decent daycare: those are not the consolation prize for people who failed to build something bigger. They are the overwhelming majority of what entrepreneurship is, they employ close to half the private-sector workforce, and building one well is a serious life's work. If this course has done its job, you now know how to tell whether the one you are considering can actually pay for itself.

Try it

A landscaping business earns 180,000 dollars of SDE. One commercial client accounts for 45 percent of revenue, the owner personally sells every job and knows every customer, and there are no written processes. What is it plausibly worth, and what would you do over two years to change that?

Answer: Every risk factor in the table points down, so this business sits at the bottom of the range or below it, plausibly around two times SDE, which is roughly 360,000 dollars, and a buyer may want a large part of that contingent on customers staying. Over two years: reduce the anchor client below 20 percent of revenue by adding accounts even at slightly lower margin; hire and train a salesperson and an operations lead so the revenue no longer depends on the owner; convert one-off jobs into annual maintenance contracts to create recurring revenue; document routes, pricing, and procedures; and clean up the books so an accountant can produce a defensible quality of earnings. If those moves lift SDE modestly to 200,000 dollars and the multiple to three, the business is worth about 600,000 dollars, a two-thirds increase driven almost entirely by reducing the buyer's risk rather than by selling more.

Common misconceptions

• "An exit means being acquired or going public." The most common good ending is that the business keeps operating and keeps paying its owner.
• "Revenue determines what a business is worth." Small businesses are priced on discretionary earnings, and the multiple reflects the buyer's risk.
• "Nonprofits cannot earn money." Nonprofit is a tax status with a nondistribution constraint; many earn most of their revenue from services.
• "Benefit corporation and B Corp are the same." One is a state legal entity form, the other a private certification.
• "Microcredit lifted millions out of poverty." Randomized trials found modest investment effects and little average effect on income, consumption, health, or schooling.
• "Starting is the only real entrepreneurship." Buying an existing business skips the survival base rate and the cash-flow gap, and is financeable in ways a startup is not.

Recap

• Most businesses end by continuing to operate or by a modest sale; public offerings are statistically negligible.
• Small firms commonly sell at two to three times seller's discretionary earnings, and the multiple rises as owner dependence and customer concentration fall.
• Nonprofit status carries a nondistribution constraint, and benefit corporations differ from B Corp certification.
• The microcredit randomized trials are the model for honest impact measurement: state the outcome, measure it, publish the disappointing result.
• Acquiring a business buys survival, cash flow, and financeability, at the cost of diligence against overpaying and owner-dependent revenue.
• Decide using opportunity cost, downside tolerance, the personal guarantee, reversibility, and motivation, and prefer the staged path.

Sources

1. U.S. Small Business Administration. (2025). Close or sell your business. sba.gov
2. Internal Revenue Service. (2025). Charities and nonprofits. irs.gov
3. Abdul Latif Jameel Poverty Action Lab. (2015). Microcredit: Impacts and limitations. povertyactionlab.org
4. Stanford Graduate School of Business, Center for Entrepreneurial Studies. (2024). Search fund study. gsb.stanford.edu
5. Wikipedia contributors. (2025). Benefit corporation. en.wikipedia.org