What's a good PMF score? The 40% benchmark, and how to read yours

It's the most common question founders ask about product-market fit, and it has a refreshingly specific answer.

The short answer: 40%

A good PMF score is 40% or higher on the Sean Ellis test. You ask your engaged users one question — "How would you feel if you could no longer use this product?" — and count the percentage who answer "very disappointed." Hit 40% and you've got product-market fit. Fall short and you don't yet, with the gap telling you how much ground is left.

That's the benchmark. Now here's why it's the benchmark, and why the single number is the start of the story, not the end of it.

Where the 40% line comes from

Sean Ellis — the marketer who coined the term "growth hacker" and ran early growth at Dropbox and LogMeIn — kept noticing the same thing across the startups he worked with. Some grew the moment you pushed on them. Others sputtered no matter how good the marketing was.

So he looked for a leading indicator, surveying users across roughly 100 startups with that one question. The pattern was clean: companies where at least 40% of users would be "very disappointed" to lose the product could grow sustainably. Below 40%, growth was a grind that usually stalled.

It's an empirical benchmark, not a law of physics. But it has held up well enough, for long enough, to become the industry standard — the closest thing startups have to a pass/fail line for fit.

How real companies score

Benchmarks get a lot more useful when you can see where actual companies landed:

• Slack — 51%. An independent survey of 731 users put Slack well above the line. Textbook fit, and the growth showed it. The full breakdown is here.
• Superhuman — 22% → 58%. The email startup started below the line at 22%, treated the score as a metric to improve, and climbed to 58% in three quarters. Proof the number is movable. Here's the engine they used.
• Baremetrics — 32%. The analytics company ran the survey and scored below 40% — and still got enormous value, mining the feedback to drive its roadmap.

Two of those three were at or above 40%. One wasn't — and used the survey anyway. Which is the whole point of the next section.

How to read your own score

The threshold is a line, but your score lands you in one of three rough zones. Treat these as guidance, not gospel:

40% and above — you have fit

Congratulations, this is the real thing. Your job shifts from finding fit to not losing it: pour fuel on growth, protect the experience that earned the score, and keep measuring so you catch any erosion early.

25–40% — you're close

This is the most common — and most workable — zone. You have a real core of people who'd miss you; you just don't have enough of them yet. This is exactly where Superhuman sat at 22%, and exactly where the score is most worth improving. Dig into your "very disappointed" users, find what they have in common, and build more of what they love.

Below 25% — rethink, don't just iterate

A low score isn't a death sentence, but it's a signal that tweaks won't be enough. The question shifts from "what feature is missing?" to "are we even solving a real problem for a specific group?" This is the moment to question the value proposition itself, not just the roadmap.

Why the number alone can lie

Here's the trap: it's easy to obsess over the percentage and miss the three things that actually determine whether it means anything.

• Who you surveyed. The score only counts if you ask engaged users — people who've genuinely experienced the product, not everyone who ever signed up. Survey your whole signup list and you'll drag in tire-kickers who were never going to care, and your score will read artificially low. This is the single most common way the number gets distorted.
• Which way it's moving. A 35% that climbed from 20% is a far healthier sign than a 42% sliding down from 55%. A single reading is a snapshot. The trend is the real signal — which is why the score is worth tracking, not just calculating once.
• The "why" behind it. Two companies can both score 40% for completely different reasons. The open-ended answers — why people would miss you, who they are, what's missing — are where the score turns into a roadmap. Slack found "video conferencing" that way. Yours are hiding in your responses.

So does the benchmark change by industry or stage? A little, at the edges — early products with tiny samples should read any single number as directional. But 40% is a remarkably durable cross-industry line, and chasing a "perfect" custom threshold is almost always a worse use of time than fixing the three things above. (Want vertical-specific guidance on how to measure? See the PMF benchmarks by industry.)

A good PMF score is 40%. A useful PMF score is one you measured on the right people, tracked over time, and read closely enough to know what to do next.