The most common way a project dies is not a technical flaw, a weak team or a lack of money. It is having built something nobody needed. Analyzing 431 post-mortems of funded startups, CB Insights found that 43% fail for that reason alone: a product with no fit to a real need. Behind that figure lies a mistake that is not commercial but philosophical: confusing one's own conviction with proof. It is the question Pieter Levels asks without naming it in MAKE (what proves that a project exists?), and his answer is blunt: a signal from outside. Everything else is conversation.
The figure that should reorder priorities
The ranking of failure causes is counterintuitive. In the analysis CB Insights published in March 2026 on 431 funded startups shut down since 2023, 70% "ran out of capital", but the study itself notes that this is almost always the final cause, not the root cause. Behind it: 43% had not found product-market fit, 29% arrived at the wrong time, 19% had unsustainable unit economics. You burn your cash trying to sell what the market was not asking for.
43% of startups die because nobody wanted their product. Running out of capital, cited in 70% of cases, is only how they find out. Building was never the problem. Building the wrong thing is.
Why a sale proves what a compliment does not
Levels puts it plainly: monetization is validation. The phrase sounds trivial; it rests on an asymmetry that economics formalized fifty years ago.
A compliment is free. "Good idea" or "I'd use that" cost nothing to the person saying them, and so carry no reliable information. A payment, on the other hand, costs something. That cost is exactly what makes it credible. Michael Spence's signaling theory (Nobel Prize in Economics 2001) establishes this in a very different context: only a signal that is costly to the sender carries verifiable information, because it cannot be imitated for free by those who are bluffing.
A sale is the costly signal par excellence. It says what no market study, no intent survey and no launch-day applause will ever say: someone chose to part with their money rather than keep it. Likes, sign-ups and congratulations are cheap signals, and therefore noisy.
What this indicates — A signal is worth only what it costs the person sending it. A sale validates because it hurts; praise does not validate because it is free.
The hard corollary: the rest is intention, not existence
If only the external signal proves anything, then everything that comes before it does not have the status we give it. A product that has been built but that nobody uses is not a product: it is an executed intention. A roadmap is not a trajectory, it is a wish list. Ten projects without a single paying user are not a portfolio but ten hypotheses. However careful it is, building does not cross the threshold of reality until an external signal has crossed it on its behalf.
This is the market version of a discipline that people who build systems know well: the presence of a thing does not prove that it works. A file exists without being read, a mechanism runs without ever firing on a real case, a dashboard shows "active" while nothing moves forward. In a market as in a machine, the proof is never in the object; it is in the usage observed outside.
A product without users, a roadmap without traction, ten projects without a sale: these are not assets, they are intentions. The threshold of reality is crossed from the outside, never from the inside.
The limit Levels himself acknowledges
A doctrine is judged by how honest it is about its blind spots. This one has two. The first, Levels states in black and white: survivorship bias. Those who fail do not write manuals about their failure. The rule "charge money, let the market decide" comes from someone for whom the market decided favorably, and nothing guarantees that the same method will reproduce the same result. The criterion is sound; the promise of success attached to it is not.
The second blind spot is subtler. The external-signal test is blind to markets that do not yet emit a signal. A right but premature idea will not make any sales, not because it is wrong, but because the market has not yet tipped. Applied without judgment, "no sales = bad idea" condemns exactly the bets that time will prove right. The external signal says what is real today; it says nothing about what will become real.
What this indicates — The external-signal criterion measures the present, not the future. It eliminates illusions; it does not settle timing. These are two separate diagnoses.
What this changes in the decision sequence
The consequence is not "charge sooner". It changes the first question a founder asks.
The instinctive question faces inward: is my idea good? is my product ready? These are conversation questions: you can answer them alone, indefinitely, without ever touching reality. The question that follows from MAKE faces outward: what would be the cheapest external signal to get, and have I already gone looking for it? A reply, a click, a first euro: anything that comes from someone other than yourself and costs them something.
This reversal is uncomfortable because it takes away the founder's right to validate themselves. That is the whole point. It replaces endless deliberation with a test that can fail, and a test that can fail is the only thing that can teach you anything. The 43% who die from a missing need did not look for it in time.
The idea is still necessary. Building is still necessary. But neither one makes a project exist. Only the outside does.
Sources
- CB Insights, Why Startups Fail: Top 9 Reasons, March 5, 2026 — 431 funded startups shut down since 2023: 70% capital ("final cause, not root cause"), 43% product-market fit, 29% timing, 19% unit economics — cbinsights.com. The 2019 edition, based on about a hundred post-mortems, reported 42% "no market need".
- Pieter Levels, MAKE: The Indie Startup Manual — readmake.com (chapters Idea, Monetize; survivorship bias warning in the foreword).
- Michael Spence, Job Market Signaling, Quarterly Journal of Economics, vol. 87, no. 3, 1973, pp. 355-374 — costly signaling theory. Nobel Prize in Economics 2001.