Product-market fit statistics: 15 numbers founders should know

Founders love to say product-market fit is a feeling. Investors, increasingly, want a number. The good news is there's real data here — some of it sobering, some of it useful as a benchmark for your own product. Here are fifteen statistics grouped into four questions: how much does fit matter, what counts as good, what did real companies score, and what does it take to measure honestly?

How much product-market fit matters

42% of failed startups fail because of "no market need." In CB Insights' widely cited analysis of startup post-mortems, this was the single most common reason for failure — essentially a fancy way of saying "no product-market fit." It consistently ranks at or near the top of every startup-failure study. It's still the #1 killer, and cheaper AI-era building hasn't fixed it.

Roughly 9 in 10 startups fail overall. The exact figure varies by source and definition, but the commonly cited estimate is that around 90% of startups don't make it — and lack of market need is the throughline in a huge share of those.

Premature scaling is a leading cause of death. The Startup Genome project found that scaling before fit — hiring and spending ahead of real demand — is one of the most common reasons startups fail. Fit isn't just a milestone; measuring it is what tells you whether you've earned the right to scale.

What counts as "good": the benchmark

40% is the product-market fit benchmark. The number comes from Sean Ellis, who benchmarked roughly 100 startups and found a pattern: those where at least 40% of users would be "very disappointed" to lose the product tended to grow sustainably, while those below it stalled. It's now the most widely used single threshold for fit.

The 40% line is a rule of thumb, not a law. Ellis himself framed it as a strong signal, not a guarantee — some products thrive slightly under it, some struggle above it. What a "good" score means also shifts by context, which is why the trend of your score matters more than any single reading.

What real companies actually scored

Slack: 51%. An independent survey of 731 Slack users found 51% would be "very disappointed" without it — well above the benchmark and a textbook case of strong fit. Here's how that measurement was done and what it teaches.

Superhuman: 22% → 58%. When founder Rahul Vohra first measured, only 22% of users would be "very disappointed" — below the line. By systematically studying his most disappointed users and converting fixable fence-sitters, he raised it to 58%, as documented in First Round Review. The engine he built to do it is copyable.

The gap between 22% and 58% was closed on purpose. That's the most important statistic on this page: a below-benchmark score isn't a verdict, it's a starting point. Scores move when you work them.

What it takes to measure it honestly

~40 responses gives a directional read; 100+ makes it reliable. Below 40 valid responses, a PMF score is too noisy to trust; by 100 it's solid enough to act on and show investors. Here's the full sample-size math.

The margin of error at 40 responses is about ±15 points. A 40% reading on 40 people could really be anywhere from ~25% to ~55%. It narrows fast: roughly ±10 at 100 responses, ±5 at ~385. Confidence grows with the square root of the sample, so the first hundred responses do most of the work.

Who you survey matters as much as how many. Sample-size formulas assume a fair draw from your real users. Survey the wrong people — inactive signups or hand-picked superfans — and you get a precise, confident, wrong number. Quality of sample beats quantity every time.

A note on these numbers

Two honest caveats. First, startup-failure percentages vary by study, definition, and year — treat "42%" and "90%" as the widely cited consensus, not precise constants. Second, a benchmark is a comparison, not a target: the goal isn't to beat Slack's 51%, it's to measure your own score, understand who it comes from, and move it up over time. A number you track beats a number you quote.