Entrepreneurship: How to Be Wrong Cheaply

Two people each have an idea for an app that helps restaurants cut food waste. One spends eight months and the savings from a previous job building it, and launches to silence. The other makes a one-page description, phones forty restaurant managers, and learns by Thursday that most of them already track waste on a whiteboard and would pay for something quite different: a tool that predicts tomorrow's orders. The second founder is not smarter. She has just found the cheapest place to be wrong.

This is the capstone of the business degree because it uses everything else: accounting to know whether a customer pays back, finance to value the company and price its shares, marketing to find the customer, operations to deliver. It adds one thing that none of them has: working out what to build when nobody can tell you.

Start with a problem someone already pays to solve

The usual mistake is to fall in love with a solution. The disciplined habit, taught almost everywhere that entrepreneurship is taught, is to start with a problem: who has it, how often, how badly, and what they do about it now. The last question matters most. A problem people already spend money, time or hacks on is real. A problem they only say is annoying may not be.

Steve Blank, whose customer development method sits under much of modern start-up teaching, put the central point as "get out of the building": facts about customers are outside, not in your head. The practical craft is asking in a way that produces evidence rather than politeness. Rob Fitzpatrick's The Mom Test (2013) gives a good rule: ask about their life and past behaviour, not about your idea. A compliment is the cheapest thing a person can give you.

What you hear	How much it tells you
"Great idea, I'd definitely use that."	Almost nothing. Costs them nothing to say.
"You should add feature X."	An opinion. Useful for ideas, weak as proof.
"Last month I spent six hours doing this by hand."	A real behaviour. Worth following up.
A deposit, a pre-order, a signed letter of intent.	The strongest signal: something was given up.

Experiments are cheap; being wrong is expensive

Eric Ries, who coined "lean start-up" in 2008 and published the book in 2011, drew on lean manufacturing and Blank's work to describe the loop: build the smallest thing that tests a belief, measure what customers do, learn, and decide whether to carry on or change direction. That smallest thing is the minimum viable product (MVP): not a bad product, but the cheapest experiment that teaches you the most. It can be a landing page, a hand-made service you pretend is software, or a video.

The method has critics. Ben Horowitz wrote in 2010 that "running lean" can become an end in itself instead of a means to winning a market. Teppo Felin, Alfonso Gambardella, Scott Stern and Todd Zenger argued that copying factory-style iteration yields only incremental ideas, and that bold start-ups need a theory, not just A/B tests. Others note that some products, such as medical devices or hardware, cannot be cheaply tested. In its favour: a randomised trial of Italian early-stage ventures (Camuffo, Cordova, Gambardella and Spina, Management Science, 2020), where entrepreneurs were taught to treat their idea as a hypothesis, found they dropped weak ideas more often, pivoted in a more focused way and earned higher revenue on average; a later multi-trial replication of 759 firms found a similar pattern with fewer pivots. The evidence supports thinking like a scientist, not any specific toolkit.

One caution is arithmetic. Small samples are noisy, and a founder who sees "3 of 40 people clicked" does not know much. If the true click rate is 10%, a test with 40 visitors can easily show 3% or 18%. Try it: choose a true conversion rate and a sample, and run twenty identical tests.

At 10% and 40 visitors the observed rate across twenty tests typically wanders between about 2% and 18%, and a single test of 4 conversions gives a 95% range from 4% to 23%. At 2,000 visitors the same test pins it within roughly a point and a half either way. The lesson: before believing a result, ask how many people it is based on.

Unit economics: does one customer pay back?

A business is a pile of customers, so the question to ask first is about one. The contribution margin is what a customer brings in per month minus the variable cost of serving them. Churn is the share who leave each month, which makes the average customer last 1 ÷ churn months. Multiply and you have the lifetime value (LTV). Against it stands the customer acquisition cost (CAC): all the marketing and sales spend per new customer.

A subscription costs $30 a month, $6 to serve, and loses 5% of customers a month. Contribution is $24, the average life 20 months, LTV $480. If acquiring a customer costs $200, LTV is 2.4 times CAC and the customer pays back in 8.3 months. A common rule of thumb in start-up circles asks for an LTV of at least three times CAC and a payback within a year; those are conventions, not laws, and they depend on how long you can wait for cash. Play with it: cut churn and watch the lifetime grow much faster than you expect.

If the price does not even cover the cost to serve, no volume rescues you, which is the break-even lesson in Accounting (34/50). If it covers it but LTV is below CAC, every customer you win loses money, and growth makes the hole bigger faster. This is the check that most business plans skip.

Runway and burn: how long you have to learn

Burn is the cash a company spends each month beyond what it earns; runway is how many months the cash lasts at that rate. A company with $250,000, spending $40,000 a month and earning $10,000, burns $30,000 and has about 8 months, if nothing changes. But revenue is rarely flat. Paul Graham's essay Default Alive or Default Dead? (October 2015) gives the sharper test: given current spending and growth, will the company reach profit before the money runs out, with no further fundraising? If yes it is default alive; if no, it is betting on investors.

With the starting numbers, cash goes negative in month 11 and dips to about −$32,000 before revenue overtakes spending in month 15. Raise growth to 15% and the company survives with a trough of about $53,000 left. The gap between those two worlds is five points of monthly growth, which is why founders sweat the growth rate and why runway should be counted in months of learning, not dollars.

Funding paths, and the power law

Money has a price, and each source charges a different one.

Path	You give up	Fits
Bootstrapping (customers fund you)	Speed	Businesses that can earn from day one
Loans (banks, government-backed schemes such as the US SBA or the UK Start Up Loans)	Fixed repayments; often a personal guarantee	Steady, predictable cash flow
Angels (individuals)	Some equity, advice	Early stage with a story but little revenue
Venture capital	A large stake, board seats, pressure to grow fast	Companies that could become very large
Grants and crowdfunding	Time spent, public exposure	Science, social missions, consumer products

Venture capital is the one that shapes the culture. Investors in a fund expect most of their companies to return little, and a few to return enormous multiples. Peter Thiel argues in Zero to One that in a good fund the best investment returns more than all the others combined; whether or not that holds in every fund, the shape is a power law: a heavy tail where a few outcomes dominate. It explains why VCs push for huge growth and "swing for the fences" even when a solid small business would be a fine life for the founder. The two goals are not the same, and the founder should know which game they are in. As a dated example, Y Combinator's standard deal (as stated on its website in November 2026) is $500,000: $125,000 for 7% of the company on a post-money SAFE plus $375,000 on an uncapped MFN SAFE.

Here is a toy fund of 20 companies where half return nothing and the rest follow a heavy-tailed distribution. The numbers are invented to show the shape, not to describe real funds.

With the default tail, the average fund of 20 returns about 1.4 times the money, but the median fund returns only 0.88 times, and 56% of funds lose money. In a typical fund the single best deal makes up around 40% of everything returned. A few huge winners carry the average, which is how a strategy can look great on average and still lose for most players.

Equity and dilution: owning less of more

Selling shares funds the company but shrinks your slice. If an investor puts in $1 million at a $4 million pre-money valuation, the company is worth $5 million post-money and the investor owns 20%; everyone else is diluted by 20%. Add an option pool for future hires, usually set aside before the money arrives so that it dilutes only the existing owners, and a second round, and the founders' share drops again. Founders do this on purpose: 28% of a $25 million company beats 100% of nothing. A SAFE (a simple agreement for future equity, created by Y Combinator) delays the price until a later round but dilutes just the same.

With the defaults, the two founders start with 100% and end with 56% between them (28% each) after the seed and Series A, while the company is worth $25 million after the Series A: $14 million on paper for the founders together. How the founders split the company between themselves matters just as much; Co-Founders: Choosing, Splitting, Staying Together (04/08) covers that.

Why start-ups fail: the sourced data

Start with the base rate, because every founder is an exception in their own mind. The US Bureau of Labor Statistics follows every new employer establishment through a cohort. For those opened in March 2013, 50.6% were still operating five years later and 34.7% after ten; about four in five survived the first year. The caveats are real: this counts new workplaces, not venture-backed start-ups; closing is not always failure (the owner may retire or sell); and the figures are for the US.

The causes are harder to measure. CB Insights collected the post-mortem notes written by founders of failed start-ups. In its 2018 report of 101 such cases, the most common cause, cited by about 42%, was "no market need": building something nobody wanted enough to pay for. Later editions found running out of cash or failing to raise more was a leading cause, in roughly 29% of 242 cases. I could only check these figures through press summaries of the report, not the report itself. They also come from founders explaining their own failures, mostly in venture-backed companies, so treat them as indicative. Taken with the experiments, unit economics and runway above, the pattern is plain: build something people want, make sure each customer pays back, and keep enough months to find out.

What a graduate just sees

• The cheapest way to be wrong is to talk to customers before building anything.
• Compliments are free; deposits, pre-orders and past behaviour are evidence.
• Small tests are noisy. Always ask how many people a result rests on.
• Look at one customer first: contribution margin, payback and lifetime value against the cost to win them.
• Runway is measured in months of learning, and "default alive" is the test of whether the plan survives without a rescue.
• VC money is for businesses that could be enormous; it suits some founders and is wrong for most.
• Every round of funding trades a slice of the company for time. Know the price in dilution.
• Most new businesses do not reach year ten. Survivors mostly learned faster than they spent.

Where it shows up next

• Product Innovation & Go-to-Market (34/63) and Pricing (34/61): how the product is shaped, priced and launched.
• Organizational Behavior (34/53) and Business Law & Ethics (34/55): hiring, co-founder disputes, company forms and what founders owe investors.
• Corporate Finance (34/51): valuing a company that earns nothing yet; the discount rates behind a venture investor's demands.
• In the world: a corner shop, a software start-up, a spin-out from a university lab, an internal team inside a large company (see Starting Something Inside a Company, 04/09).

If you want to go deeper

• 📘 Textbook. Bill Aulet, Disciplined Entrepreneurship: 24 Steps to a Successful Startup (Wiley, 2013). The step-by-step method taught at MIT.
• 📘 Textbook. Steve Blank and Bob Dorf, The Startup Owner's Manual (K&S Ranch, 2012). The customer-development process in detail.
• 📘 Gentler. Rob Fitzpatrick, The Mom Test (2013), on asking questions that give real answers; Eric Ries, The Lean Startup (Crown, 2011); and Paul Graham's essays at paulgraham.com.
• 🌐 Free course. MIT OpenCourseWare 15.390 New Enterprises (Spring 2013, instructors William Aulet, Howard Anderson and Matt Marx): the full course with readings and assignments.