Someone will tell you this week that 90% of startups fail. A LinkedIn post, an accelerator pitch, a well-meaning relative at dinner. It's the most repeated statistic in entrepreneurship, and here's the strange part: almost nobody who repeats it can tell you where it comes from. I went looking. The trail runs cold in a misread report from 1975, a circular citation chain, and a definition of "startup" that probably doesn't include you. The claim that 90% of startups fail isn't a measurement. It's folklore with a percentage sign attached. And believing it changes how founders behave, usually for the worse.
Let's pull the thread.
Do 90% of startups really fail?
The short answer is no, not by any general definition of startup or failure. The real number ranges from roughly 20% to 90% depending on what you count as a startup, what you count as failure, and how long you wait before counting. That range isn't a technicality. It's the whole story.
If "startup" means any new business and "failure" means closing down, the US government's own data says about 1 in 5 die in year one. If "startup" means a venture-backed company chasing a 10x return and "failure" means investors didn't get their money back, the number climbs to 75%. Only when you narrow the definition all the way to "scalable tech startups attempting venture-scale outcomes" does anything close to 90% show up, and even then it measures failure to hit a specific financial bar, not failure to build a real business.
So when someone quotes 90% at you, the right response is a question: 90% of what, failing at what, by when? Almost nobody who cites the number can answer.
Where did the 90% statistic come from?
Nobody can produce a primary source, which is the tell. Researchers who've traced the claim keep hitting dead ends. One trail leads back to a 1975 Dun & Bradstreet report on business failures that never actually claimed a 90% failure rate; journalists misread data about the age of companies that failed in a single year, and the number entered circulation. It's been repeated for five decades since, mostly without anyone checking.
The modern trail is just as shaky. Recent versions of the claim often cite a Startup Genome report, which in turn cites a Small Biz Trends article that researchers can't locate. The Small Business Administration, which gets credited with the statistic constantly, has pushed back on it because its own data shows nothing of the sort.
Think about that. The most quoted number in startup culture is a citation loop with no floor. If a founder presented market sizing built like this, any decent investor would walk.
What do the real numbers actually say?
Verified data tells a much more specific story: about 20% of new businesses fail in year one, about half survive five years, and roughly a third make it to ten. Those figures come from the Bureau of Labor Statistics, which tracks every new private-sector establishment in the country. The latest numbers put the one-year failure rate at 20.4%, the five-year rate at 49.4%, and the ten-year rate at 65.3%.
The numbers get worse as the definition gets narrower:
| Population | Failure definition | Rate |
|---|---|---|
| All new US businesses (BLS) | Closed within 1 year | 20.4% |
| All new US businesses (BLS) | Closed within 5 years | 49.4% |
| All new US businesses (BLS) | Closed within 10 years | 65.3% |
| Software/information sector (BLS) | Closed within 10 years | 70.9% |
| Venture-backed, $1M+ raised (Harvard) | Never returned investor cash | 75% |
| Scalable tech startups (Startup Genome) | Missed venture-scale returns | ~90% |
The 75% figure comes from Harvard Business School researcher Shikhar Ghosh, who studied about 2,000 companies that raised at least $1 million between 2004 and 2010. Three quarters never returned cash to investors. But note what that measures: 30 to 40 percent of those companies liquidated with total losses. The rest "failed" by returning less than the fund hoped while often still operating as real businesses. A company doing $2 million a year in revenue that never pays back its VCs is a failure in Ghosh's data and a wild success by any bootstrapper's math.
And that ~90% figure from Startup Genome? It applies to scalable, innovative startups measured against venture-scale outcomes. It was never a claim about new businesses in general. The number is real in its narrow lane. The way it gets quoted is not.
If you want the full breakdown by industry, stage, and timing, I've gone deep on it in our startup failure rates data guide. This piece is about the myth itself.
Why does the myth refuse to die?
The 90% number survives because everyone who repeats it gets something out of it. That sounds cynical, so let me be specific about the incentives.
Accelerators and VCs benefit from a scary baseline. "90% fail, but our portfolio companies beat the odds" is a sales pitch, and the worse the baseline, the better the pitch. Content marketers benefit because fear outperforms nuance; "90% of startups fail" gets clicks that "49.4% of businesses close within five years, for heterogeneous reasons" never will. Course sellers and gurus benefit for the same reason insurance salesmen mention house fires.
And founders repeat it too, for a subtler reason: it's flattering. If 90% fail, then merely surviving makes you a statistical marvel, and failing puts you in the overwhelming majority. The myth offers drama on the way up and absolution on the way down. That's a hard product to compete with.
There's also a simpler mechanism. A number that specific sounds measured. Round, dramatic, easy to remember. The perfect meme. Accuracy was never part of its fitness function.
What does believing the myth cost founders?
Bad odds produce bad strategy, in two opposite directions. I've watched both happen.
The first failure mode is lottery-ticket thinking. If failure is nearly certain anyway, why bother with discipline? Founders in this mode skip customer conversations, skip the financial model, skip pricing research, and sprint straight to building, because the whole thing is a moonshot and moonshots are about speed and luck. Except the data says the opposite. When CB Insights analyzed 431 failed VC-backed companies, 43% cited poor product-market fit as a root cause. 70% ran out of cash, but CB Insights classifies that as the final symptom, not the disease. These companies didn't lose a lottery. They built things nobody wanted badly enough to pay for, which is the single most preventable cause of death a startup has.
The second failure mode is not starting at all. Plenty of would-be founders with viable, modest ideas (a niche SaaS tool, a service business, a productized consultancy) look at "90%" and keep the day job. But their actual odds were never 90% against. For a bootstrapped small business, the five-year picture is close to a coin flip, and the founder's choices weight the coin heavily.
Both failure modes come from the same error: borrowing the odds of a population you don't belong to.
Which failure numbers actually apply to you?
Match the dataset to your situation, because the difference is enormous. A few common cases:
Bootstrapping a small SaaS or service business to a few thousand a month? The BLS numbers are your baseline: about 80% survive year one, about half reach year five. In the software-heavy information sector the ten-year picture is tougher, with 70.9% closing within a decade, but that's still a long way from 9-in-10 doom.
Raising venture capital to chase a big outcome? Ghosh's 75% is your honest reference point, with the caveat that "failure" there includes companies that lived but didn't return the fund. If you take VC money, you've signed up for venture math, and venture math is brutal on purpose.
Building the next breakout, blitz-scaling tech startup? Fine, the 90% figure is roughly yours. You've chosen the hardest game in business. At least you're quoting the right statistic.
The point isn't that the odds are secretly great. Half of businesses closing within five years is sobering. The point is that your odds are conditional, and the conditions are substantially under your control.
What should you do with the real odds?
Treat failure rates as a list of preventable causes, not a prophecy. The data is remarkably consistent about what kills companies: building something the market doesn't want, running the bank account down before finding out, and unit economics that never worked. Every one of those has a countermeasure that costs weeks, not years.
Before you build, run real validation: 20+ customer discovery interviews, a landing page test, a pre-sale if you can manage it. Before you spend, build even a crude financial model so you know your runway and your break-even point. Write down your riskiest assumption and design the cheapest possible test for it. None of this is glamorous. All of it moves you out of the failure columns that dominate the data.
You can do this work in a spreadsheet and a Google Doc. Plenty of founders manage it in Notion. If you want more structure, a planning tool like Foundra walks first-time founders through validation, financial projections, and go-to-market step by step, and there's a set of free startup calculators (runway, startup costs, market sizing) that cover the math pieces on their own.
Whichever route you take, the founders who beat the averages are mostly the ones who checked whether the market wanted the thing before betting everything on it. That's it. That's the edge hiding inside all these statistics.
Key takeaways
- No general dataset supports "90% of startups fail." The claim traces to a misread 1975 Dun & Bradstreet report and survives on circular citations.
- Verified BLS data: 20.4% of new US businesses fail in year one, 49.4% within five years, 65.3% within ten.
- Harvard's Shikhar Ghosh found 75% of venture-backed startups never return investor cash, but only 30 to 40 percent lose everything.
- The ~90% figure is real only for scalable tech startups measured against venture-scale returns. It was never about new businesses in general.
- The myth persists because it sells: scary baselines flatter accelerators, VCs, content marketers, and even founders.
- Believing it produces two errors: reckless lottery-ticket thinking or never starting. Both come from borrowing another population's odds.
- The top preventable cause of failure is building something nobody wants (43% of failures in CB Insights' data). Validation is the countermeasure.
FAQ
Is it true that 90% of startups fail?
Not by any general definition. About 20% of new US businesses fail in year one and about half within five years, per BLS data. The 90% figure only applies to scalable tech startups measured against venture-scale financial returns.
Where did the 90% failure statistic come from?
No primary source exists. Researchers trace it to a misinterpreted 1975 Dun & Bradstreet report, and modern citations loop through a Startup Genome report referencing an article nobody can find. The SBA has disputed the claim.
What percentage of small businesses fail in the first year?
20.4% of new US private-sector establishments close within their first year, according to the Bureau of Labor Statistics. In other words, about 4 in 5 survive year one.
What is the failure rate for venture-backed startups?
About 75% never return cash to investors, based on Shikhar Ghosh's Harvard study of roughly 2,000 companies that raised $1 million or more. Total wipeouts are rarer: 30 to 40 percent of those companies.
Why do most startups actually fail?
Building something the market doesn't want. CB Insights found 43% of failed startups cited poor product-market fit as a root cause, and classifies running out of money (70%) as the final symptom rather than the underlying cause.
Does the 90% number apply to bootstrapped businesses?
No. Bootstrapped businesses match the general BLS survival curve: roughly a coin flip over five years, with odds that improve materially with validation, cash discipline, and working unit economics.
Top comments (0)