How to Measure Product Market Fit (Beyond the 40% Rule)

Sean Ellis arrived at the 40% threshold empirically, surveying hundreds of high-growth startups and finding that companies above that mark grew consistently faster than those below it. The number is not arbitrary — it represents a meaningful inflection in how strongly a customer segment depends on a product. Below 40%, dependency is optional. Above it, the product begins to approach irreplaceable for that segment.

The limitation is that the test is attitudinal, not behavioural. It captures how users feel about losing the product — not whether they actually use it in the way that would create that dependency. A user who logs in occasionally and answers 'very disappointed' counts the same as a user who runs core business workflows through the product daily. The aggregate percentage hides who is actually generating the PMF signal and who is inflating it.

The deeper problem: the test cannot tell you why you are at 30% instead of 40%, or which dimension to improve to close the gap. Is the score low because the problem isn't urgent enough for your segment? Because the product doesn't solve it completely? Because the wrong customers are in the survey? Because execution is too slow? These are four completely different root causes requiring four completely different remedies. A single question cannot distinguish between them.

Market readiness: whether the customers you are targeting have the problem urgently enough, understand it clearly enough, and have the budget and authority to pay for a solution. A technically excellent product targeting a market that doesn't recognise the problem will score below 40% not because the product is wrong but because the market isn't ready to receive it — and that requires measurement separate from product quality.

Execution readiness: whether the team's processes, priorities, and decision-making patterns are aligned with closing the PMF gap. A product with strong PMF signal can fail to convert that signal into growth because of execution leakage — misaligned roadmap priorities, customer success gaps, or founder decision-making patterns that consistently delay the actions that would move the needle. You cannot detect this with a customer survey.

Founder readiness: whether the founder's cognitive patterns are distorting how they read the data. Confirmation bias causes founders to interpret a 38% score as 'almost there' when it may be measuring the wrong user segment entirely. Overconfidence causes founders to act on the 40% number without segmenting it by ICP or customer vintage. The measurement of PMF requires measuring the measurer — which no single-question survey does.

A comprehensive PMF measurement approach scores three broad categories: market evidence (the quality and specificity of validation that the problem is real and urgent for a defined segment), product evidence (retention cohorts, usage patterns, replacement test responses, Sean Ellis scores), and execution evidence (how well the team responds to the data it has and how quickly it converts findings into action).

Within product evidence, retention cohort shape is more predictive than the Sean Ellis percentage alone. An early cohort that flattens above zero — even at 15–20% — is more promising than a 38% Sean Ellis score from a flat retention curve, because retained users are demonstrating dependency through behaviour rather than attitude. The cohort shape also reveals which user type is retained, allowing you to identify the segment with real PMF inside a broader mediocre average.

The value of measuring across all three dimensions is that the result produces specific, actionable findings rather than a number that requires generous interpretation. A score showing strong product evidence but weak market evidence says: the product works for the customers who have it, but you are not reaching the right customers at scale. A score showing strong market evidence but weak execution evidence says: you have the signal, but the team's patterns are preventing conversion to growth. Each finding has a different fix.