TL;DR: Being in a good market with a product that can satisfy that market. The only thing that matters for a startup’s survival. Vivid as a feeling, measurable via the Sean Ellis 40% test.

The origin

Marc Andreessen crystallized the term in 2007 in the most consequential blog post of the modern startup era (only-thing-that-matters). His claim was deceptively simple: every startup has three core elements — team, product, and market — and market matters more than the other two combined. A great market pulls product out of the startup. A terrible market kills the best product and the best team you’ll ever assemble.

Andy Rachleff’s law (the rule Andreessen credits the formulation to):

When a great team meets a lousy market, market wins. When a lousy team meets a great market, market wins. When a great team meets a great market, something special happens.

Andreessen’s working definition is the one everyone still quotes: “Product/market fit means being in a good market with a product that can satisfy that market.” You can feel when it’s happening — customers are buying as fast as you can make it, money is piling up, you’re hiring as fast as you can, reporters are calling. You can also feel when it isn’t — deals don’t close, word of mouth is silent, press reviews are lukewarm, you’re explaining the product on every call.

A startup’s life divides cleanly into BPMF (before product/market fit) and APMF (after). When BPMF, you focus obsessively on getting there. Do whatever is required: change people, rewrite the product, move to a different market, raise that fourth round. When APMF, almost any other mistake becomes recoverable.

Measuring it

Andreessen’s description is vivid — and almost useless if you haven’t gotten there yet. By the time investment bankers are staking out your house, you already have PMF. You don’t need a definition; you need a leading indicator.

Rahul Vohra at Superhuman cracked this in 2018 (superhuman-pmf-engine). Working from research by Sean Ellis (early growth at Dropbox, LogMeIn, Eventbrite), Vohra reduced the entire question to a single survey item: “How would you feel if you could no longer use the product?” Measure the percent who answer “very disappointed.”

The magic number is 40%. Below it, companies almost always struggle to grow. Above it, traction. Superhuman started at 22%, segmented to find their best users, ran a four-step optimization engine, and reached 58% within three quarters:

  1. Segment to find your supporters. Narrow the market to where you’re already loved. Build a high-expectation customer profile.
  2. Analyze feedback. Why do lovers love you? What holds fence-sitters back? Politely ignore the users who wouldn’t be disappointed.
  3. Split the roadmap 50/50. Half deepening what lovers love. Half removing barriers for fence-sitters.
  4. Track relentlessly. Make the “very disappointed” percentage the team’s most important metric.

The argument

Retention is the real signal; growth can be faked. You can buy growth through ads. You cannot buy retention. If customers stay, use the product regularly, and renew without heavy persuasion, you have PMF. Cohen puts it bluntly: “growth on a leaky bucket isn’t growth, it’s waste” (pmf-roadmap). Duolingo proved the point by making CURR (current user retention rate) their north star before layering on acquisition (duolingo-growth).

PMF with early adopters ≠ PMF with mainstream. This is the trap that has killed more once-promising startups than any other. Early adopters forgive rough edges because they’re solving a burning problem. Mainstream customers want polish, support, and proven value. The adopter-mainstream gap is real and invisible until you hit it (crossing-the-chasm-concept). Most startups fail here, not at the initial PMF stage.

PMF in too small a market isn’t PMF for a venture-scale outcome. Gumroad had real PMF within its niche — creators absolutely loved the product. The niche was just too small to support venture scale. The failure wasn’t product quality; it was market size (gumroad-failure). This connects directly to Andreessen’s core insight (market matters most) and to invisible-asymptotes — the growth ceilings you cannot see until you smash into them.

The Hook Model reveals the mechanics. Habit-forming products have PMF; transactional products need relentless growth to compensate for churn. The hook-model (trigger, action, reward, investment) is a framework for building products that retain through repetition rather than persuasion. Products with strong PMF almost always have strong hooks underneath.

Case study

For the cleanest real-world demonstration, see chatgpt-pmf. ChatGPT became the fastest-growing consumer app in history — a hundred million MAU in sixty days — because OpenAI accidentally released a research preview into a market that had been waiting two and a half years for the right packaging of GPT-3. The market did not care that the product was rough. It pulled the product out of the startup, exactly as Andreessen described. Every claim on this page has a ChatGPT-shaped piece of evidence behind it.

Loose threads

  • How does PMF interact with counter-positioning? A product can have PMF and an incumbent response problem — or PMF because the incumbent can’t respond.
  • The 40% test measures individual attachment but not market size. You need both. Superhuman’s segmentation step partly addresses this.
  • Beware novelty-effects — early engagement often decays as curiosity fades. The Ellis test partially controls for this by filtering to active users, but novelty-driven PMF is the most common false positive in the wild.

What links here