Most startups fail, and the reasons they actually fail are not the reasons their founders fear. Founders fear that someone will steal the idea, or that a bigger competitor will crush them. The record says they mostly die of something quieter, that not enough people wanted the thing they built, and that the money ran out before they found the people who did. The enemy of a startup is indifference far more often than imitation. This article sets out what the base rates and the post-mortems actually say, because a founder who knows how startups really fail can spend scarce effort guarding against the real risks rather than the imagined ones. The companion patent series made the same point from the other side, that founders overrate patents and the protection they offer. This series asks what actually decides survival. The treatment is general, and it is information rather than business advice.

A Brief History

The study of why startups fail is younger than the startups themselves. For a long time failure was treated as private and shameful, its lessons buried with the company. Two developments brought it into the open. The first was the practice, associated with the lean-startup movement and with customer development, of treating a startup as a series of testable guesses rather than a plan to be executed, which made failure a source of data rather than only of embarrassment. The second was the rise of public post-mortems, in which founders wrote candidly about why their companies died, and the compilation of those accounts into a rough statistical picture. That picture is imperfect, because it rests on self-report, but it is the best evidence available, and its central finding is stable.

The Base Rate

The first fact is the base rate, and it is sobering. The exact figure depends on how one defines a startup and a failure, so the honest statement is a range. Of new businesses broadly, roughly half are gone within five years and only a third survive ten, on long-running labor-statistics data. Of venture-backed startups, which take bigger risks for bigger outcomes, the great majority never return their investors’ capital, and estimates of outright failure commonly run from two-thirds to nine-tenths. However the line is drawn, failure is not the exception for a startup. It is the base case, and any honest reasoning about a venture begins from that fact rather than from the founder’s hope.

These numbers admit a simple model. If startups failed at a constant yearly rate, the fraction surviving to year $t$ would follow an exponential survival function,

\[S(t) = e^{-\lambda t},\]

where $\lambda$ is the yearly hazard of failure. Fitting the curve to the five-year figure, $S(5) = 0.5$, gives $\lambda = \tfrac{\ln 2}{5} \approx 0.139$ per year. That same constant rate would then predict a ten-year survival of $S(10) = e^{-1.39} \approx 0.25$, yet roughly a third of businesses survive ten years. The model under-predicts the survivors, which means the real hazard is not constant but falls as a company ages. A startup is most fragile when it is young, the liability of newness familiar to students of organizations, and the danger concentrates in the early search for a market that the rest of this series examines.

The Reasons They Actually Fail

When founders of failed startups are asked what happened, their accounts cluster into a familiar set of causes. The single most common, appearing in something like a third to two-fifths of post-mortems, is some version of no market need, the absence of product-market fit, the discovery that the product solved a problem too few people had or cared enough to pay to solve. Running out of cash and failing to raise the next round come close behind, though these are often the financial face of the same problem, since a company the market wanted can usually raise. The wrong team, being outcompeted, pricing and cost problems, a product users did not enjoy, and the absence of a business model fill out the list.

The pattern in the list matters more than any single entry. Almost every leading cause is a failure to make something a market actually wanted, or to reach and charge the people who did. These are the failures the rest of this series examines, the funnel a startup must pass through, the search for product-market fit, and the distribution that reaches paying customers.

The Fear and the Reality

Set the list of real causes against the list of founder fears, and the mismatch is stark. Founders worry that a competitor or a contractor will steal the idea and win with it. That cause is essentially absent from the post-mortems. Startups are not, in any meaningful number, killed by idea theft. They are killed by indifference, by building something that the world declined to want. An idea taken and executed by someone else is so rare a cause of death that planning around it while neglecting the market is a precise inversion of the real risks. This is why the patent series counseled against over-investing in protecting an idea. The idea was never the scarce thing. A market for it was.

The Base Rate and the Confident Founder

Nearly every founder believes their startup will be the exception, and the believing is itself a problem, because it is nearly universal, and a signal that everyone emits carries almost no information.

Let $p$ be the base rate of success, the fraction of startups that reach a good outcome, which is small. A founder’s confidence is a signal, but consider how common it is. Write $P(C \mid S)$ for the chance a founder is confident among those who will succeed, and $P(C \mid F)$ for the chance among those who will fail. Both are close to one, because the founders who failed were, at the start, just as sure as the founders who won. Bayes’ theorem gives the probability of success for a confident founder,

\[P(S \mid C) = \frac{P(C \mid S)\, p}{P(C \mid S)\, p + P(C \mid F)\,(1 - p)}.\]

When the two confidence rates are equal, the expression collapses to $p$ exactly. Confidence then tells you nothing, and the honest estimate of your odds is the base rate itself.

A worked instance. Take a base rate of success of one in ten, $p = 0.1$, a ninety-five percent confidence rate among future successes, and a ninety percent rate among future failures. Then

\[P(S \mid C) = \frac{(0.95)(0.1)}{(0.95)(0.1) + (0.90)(0.9)} = \frac{0.095}{0.905} \approx 0.105.\]

The founder’s certainty moves the true probability of success from one in ten to about one in nine and a half, almost nothing.

The mechanism is clearest in odds form. The theorem multiplies the prior odds of success by the ratio of the two confidence rates,

\[\frac{P(S \mid C)}{P(F \mid C)} = \frac{P(C \mid S)}{P(C \mid F)} \cdot \frac{p}{1 - p}.\]

That ratio, the Bayes factor, is here $0.95 / 0.90 \approx 1.06$, so confidence improves a founder’s odds by about six percent in relative terms and nothing more. A signal worth heeding would carry a Bayes factor far from one, and a founder’s certainty does not.

This is the arithmetic of base-rate neglect, and the overconfidence it feeds is itself a cause of failure, because a founder who mistakes confidence for evidence overinvests in a venture whose real odds the confidence never improved. The remedy is not to abandon the venture, since someone does clear the base rate. It is to act as though the base rate is true, to build cheaply, test early, and treat the market’s verdict, rather than one’s own conviction, as the evidence that counts.

Failure Is the Default

The base rate and the post-mortems point at the same structural truth. A startup succeeds only if many independent things go right, a real problem, a product that solves it, customers who can be reached, a price they will pay, a team that can build and sell, and enough money to last until those align. Each is a hurdle, and clearing all of them is far less likely than clearing any one. Failure is therefore the default, not because founders are foolish, but because success is a conjunction of many uncertain events. The next article makes that conjunction precise, and the articles after it take the hurdles one by one.

Epistemic State

The base rate and the cause distribution are empirical, and they rest on imperfect evidence. Post-mortems are self-reported, and a founder’s account of what killed the company can rationalize or simplify, so a cited cause such as no market need may itself mask an execution or distribution failure underneath. The figures vary with the definition of a startup and of failure, which is why this article gives ranges rather than single numbers. Survivorship bias shapes much of the popular writing on startups, which studies winners and infers lessons that the larger population of losers would contradict.

The Bayesian argument is a simplification offered for its shape, not a measurement. It assumes confidence is nearly as common among failures as among successes, which is the realistic case but not a measured constant, and it treats success as a single binary event.

The qualitative conclusion, that startups die of indifference far more than of theft, is well supported by the available record. Throughout, this is general information, and it is not business advice.

Out of Scope

The detailed taxonomy of failure causes, the specific datasets and their methods, and the variation in failure rates by sector and geography are left to the empirical literature. The psychology of founders, beyond the base-rate point made here, is a large adjacent subject. The remedies, namely how to find product-market fit, how to build under uncertainty, how to reach customers, and where durable advantage actually comes from, are the subjects of the articles that follow and are not treated here.

Conclusion

Startups fail mostly for want of a market, not for want of secrecy, and the founder who internalizes this spends differently. The base rate of failure is high, a founder’s confidence does little to change it, and the cause of death is almost always that not enough people wanted the product or could be reached and charged for it. Idea theft, the fear that looms largest, is the cause that matters least. The rest of this series follows the path a startup must walk to beat the base rate, beginning with the funnel of failure it must pass through and ending with what it takes to succeed and where moats actually come from.

References