Product-market fit: what it means, how to measure it, and what to do when you are not there

What does product-market fit actually mean?

The phrase owes its fame to Marc Andreessen's 2007 essay The only thing that matters, which defines it as being in a good market with a product that can satisfy that market, and credits the idea to Andy Rachleff, then at Benchmark Capital. Andreessen's argument is that the market matters more than the team or the product:

He also describes what it feels like. Without fit, customers are not quite getting value, word of mouth does not spread and sales cycles drag. With it, customers buy as fast as you can deliver and hiring cannot keep up. Useful, but founders are famously bad judges of their own traction, which is why a number helps.

How does the Sean Ellis 40% test work?

You ask existing users how they would feel if they could no longer use the product. Sean Ellis, who ran early growth at companies including Dropbox, set out the test in a post on his blog, The Startup Pyramid, already online by early 2010. In his experience, product-market fit requires at least 40% of users to answer “very disappointed”.

Read the original and two caveats stand out. Ellis calls the threshold somewhat arbitrary and says he set it by comparing results across nearly 100 startups: those struggling for traction were always below 40%, most of those with strong traction were above. And the very disappointed users must add up to a market large enough to build a real business.

What should the product-market fit survey ask?

These are the four questions Vohra describes, laid out as we use them; the “I no longer use it” option is our addition, to keep lapsed users out of the score. Send them together, in this order, with no warm-up that hints at the answer you want.

1. How would you feel if you could no longer use the product? Very disappointed · Somewhat disappointed · Not disappointed · I no longer use it
2. What type of person do you think would benefit most from it? (open answer)
3. What is the main benefit you get from it? (open answer)
4. How could we improve it for you? (open answer)

Read question two only among the very disappointed: it describes your best customer in their own words. Question three gives you the line for your landing page.

Which other signals confirm product-market fit?

Behavioural ones, because they do not depend on what people say.

• Cohort curves that flatten. Andreessen Horowitz looks for retention in each cohort to stabilise after roughly 6 to 12 months, and for newer cohorts to beat older ones. If you bill through Stripe, its Billing analytics already shows subscriber and revenue retention by cohort.
• Falling early churn. Fewer customers leaving in their first months is the clearest sign the product has become a habit; see our guide to SaaS churn rate.
• Demand you did not pay for. Customers arriving through word of mouth rather than ads, and sales cycles that get shorter.
• Efficient growth. David Sacks argues that needing to burn 3x or more of net new ARR to grow suggests product-market fit is not quite what it appears.

Missing fit is expensive: in CB Insights' 2026 analysis of 431 VC-backed shutdowns, 43% of the companies with stated reasons cited poor product-market fit. Once the signals line up, the dashboard changes too, and revenue metrics such as MRR, churn and CAC take over; the SaaS metrics that matter explains which ones and when.

What should you do if you score below 40%?

Keep burn low and point everything at raising the share of very disappointed users. That is Ellis's own advice, with a pointed warning: this is not the moment to hire senior marketing and sales leaders to fix it, because they add cost and are unlikely to solve it better than the founders talking to users.

• Interview the very disappointed. What convinced them, what they would use instead, what they tell colleagues.
• Sharpen the promise. Sometimes leading with the benefit they mention most is enough.
• Narrow the segment. Being essential to a few beats being liked by many.
• Change the product when needed. Ellis notes it often takes significant revisions, and occasionally a restart of the vision.
• Re-validate the problem. If you doubt the problem itself, go back to validating the idea before writing more code.

And avoid the usual ways of fooling yourself: surveying every sign-up including people who never came back, surveying only fans, averaging segments that behave differently, or treating 40% as a law rather than a reference point.