Product-Market Fit (often shortened to PMF) is the moment a startup transitions from "trying to find a market" to "trying to serve a market that has found it." Marc Andreessen popularized the phrase in his 2007 essay The Only Thing That Matters, where he argued PMF is the single binary that determines startup outcomes — every other variable (team, idea, execution) matters far less.

The shortest working definition: PMF means you're in a good market with a product that can satisfy that market. Both halves matter equally. A great product in a non-existent market is not PMF. A mediocre product in a hungry market often is — which is why teams test for fit with an MVP (Minimum Viable Product) long before they polish anything.

How to measure product-market fit: 3 standard tests

There is no single number that proves PMF. Three tests are commonly used, usually in combination.

Test	Threshold	What it measures
Sean Ellis test	>40% say "very disappointed" if product disappeared	Subjective — emotional dependence
Retention curve	Flat after 4–8 weeks (not decaying to zero)	Behavioral — do users stick
Organic growth ratio	Word-of-mouth contributes >25% of new users	Distribution — do users tell others

The Sean Ellis test — "How would you feel if you could no longer use [product]?" with options Very Disappointed / Somewhat Disappointed / Not Disappointed — was the first popularized quantitative PMF signal. Hitting 40% Very Disappointed is the threshold above which PMF is reliably present. Superhuman published the canonical playbook for iterating onto this score in 2018.

The PMF Trifecta: why passing 1 of 3 tests means nothing

Most teams pick the test that's easiest to run and call it PMF. That's the most common mis-diagnosis we see across the startups we cover on Framework — and it's why so many "PMF announcements" age badly.

The three tests above measure different things, and PMF is the state where all three signal positive at the same time:

Test	What it proves	What it misses on its own
Sean Ellis ≥ 40%	Emotional dependence (loved)	Doesn't prove repeat usage
Flat retention curve after 4–8 weeks	Behavioral stickiness (used)	Doesn't prove acquisition will be cheap
Organic share ≥ 25% of new users	Distribution leverage (told)	Doesn't prove the love is durable

A product can pass any one test and still fail the market:

Quibi (2020) had millions of downloads (acquisition) but retention collapsed inside 90 days — passing the marketing test, failing the retention test. Worth ~$1.75B raised; shut down 6 months after launch (The Information's post-mortem).
Many Series A SaaS show flat retention on a tiny base (test 2 passes) without organic share or emotional dependence — the company looks healthy until paid acquisition catches up and CAC payback collapses.
Superhuman (2017–2019) is the canonical all-three-pass case: Rahul Vohra's First Round write-up shows the team explicitly iterated until 58% Sean Ellis and flat retention and measurable referral pull. That's the bar.

The diagnostic rule: if you can only quote one test, you haven't reached PMF — you have a candidate signal. Re-test the other two before you raise, scale, or hire against the assumption.

Signs you have product-market fit

Organic growth — users tell others without being paid to
Retention curves that flatten rather than decaying to zero (the "smile curve" pattern)
The team is overwhelmed by demand rather than chasing it — support tickets pile up, servers struggle
Sales cycle compresses — prospects close themselves rather than needing convincing
Pricing power emerges — you can raise prices and not lose users

Signs you don't have it

The inverse — and these are easy to ignore because they often co-exist with positive vanity metrics:

Usage requires constant marketing push (turn off ads → growth stops)
Retention curves trend toward zero regardless of feature improvements
Customer acquisition cost trends up over time
The team spends more time selling than serving
Customer feedback is generic ("nice to have") rather than urgent ("when can I get more?")