Most teams that think they're searching for product-market fit are actually searching for permission to stop searching. The signs feel close: a few delighted customers, a flattering tweet, a renewal that came in on time. Then growth stalls and nobody can say why. The hard truth is that fit is binary in a way most metrics aren't, you either have a market pulling product out of you, or you're pushing. This guide is about telling the difference, on purpose, before you scale a business that the market hasn't actually voted for.

The quick version

  • Definition. Product-market fit means being in a good market with a product that can satisfy it. You can feel it: usage grows faster than you can keep up, and customers refer other customers without prompting.
  • It comes in a sequence. Prove the value hypothesis (someone genuinely needs this) before you chase the growth hypothesis (how to reach more of them). Scaling first is the classic way to die.
  • You can measure the feeling. Ask users how they'd feel if they could no longer use the product. If 40%+ say "very disappointed," you likely have fit. Below that, you don't yet.
  • The move. Stop guessing whether you have fit. Run the survey, find the segment that already loves you, and build harder for them rather than blandly for everyone.

The idea in depth

The phrase belongs to Andy Rachleff, co-founder of Benchmark Capital and later Wealthfront, who built it on an observation from his venture career: when a great team meets a lousy market, the market wins. It reached the wider world through Marc Andreessen's 2007 essay "The Only Thing That Matters", where he set the definition that still gets quoted: "Product/market fit means being in a good market with a product that can satisfy that market." Plain, almost boring, and that plainness is the point. Fit is not about how clever the product is. It's about whether the market wanted it.

Andreessen's more useful contribution was teaching people what fit feels like from the inside. Before it: "the customers aren't quite getting value out of the product, word of mouth isn't spreading, usage isn't growing that fast, press reviews are kind of 'blah,' the sales cycle takes too long, and lots of deals never close." After it: "the customers are buying the product just as fast as you can make it, or usage is growing just as fast as you can add more servers." The second state is unmistakable once you're in it. The trap is convincing yourself you're there when you're not.

"You can always feel when product/market fit isn't happening.", Marc Andreessen, 2007

So treat fit as a state you diagnose, not a milestone you declare. Before any big bet on hiring, paid acquisition or a sales team, ask the honest version of the question: is the market pulling, or am I pushing? If every new customer takes a heroic sales effort and none of them bring a friend, you're pushing. Pouring money on a push just buys you a more expensive version of the same problem.

Value first, growth second, the order matters

Rachleff's sharpest framing is that fit decomposes into two hypotheses, and the order is not negotiable. The value hypothesis is the bet that a specific group of people genuinely needs what you've built, the what, the who, and the why-they'd-use-it. The growth hypothesis is the separate bet about how you'll reach more of them economically. Product-market fit means you've proven the value hypothesis. Only then is the growth hypothesis worth touching.

This sounds obvious and is violated constantly. A team gets a promising early signal, raises money against it, and immediately hires marketers and salespeople to grow, before the value hypothesis is actually proven. Growth machinery applied to a product the market doesn't yet need doesn't manufacture demand; it manufactures churn, and burns the runway you needed to find fit in the first place.

flowchart TD
    A(["Value hypothesis
Does a specific group
genuinely need this?"]) -->|Proven| B(["Product-market fit
the market starts pulling"])
    A -->|Not yet| C(["Keep iterating:
narrow the segment,
sharpen the product"])
    C --> A
    B --> D(["Growth hypothesis
How do we reach
more of them, economically?"])
    C -.->|Scaling here| E(["Expensive churn
burning runway"])
Prove value before you chase growth; scaling the dotted path is the classic failure mode. Leaders Loop

The practical fix is to put a gate between the two phases and refuse to cross it on vibes. Write your value hypothesis as one sentence, "[this specific person] will use [this] because [this need] isn't met today", and decide in advance what evidence would prove it. Until that evidence exists, every dollar goes into finding fit, not scaling it.

Making the feeling measurable: the 40% test

"You'll feel it" is true but unhelpful when you're staring at an ambiguous dashboard at 9pm. The most-used answer comes from Sean Ellis, who turned the feeling into a single survey question: "How would you feel if you could no longer use [product]?" with three options, very disappointed, somewhat disappointed, not disappointed. Across roughly a hundred startups he found a threshold: products where 40% or more of users said "very disappointed" tended to grow sustainably; those below almost always struggled. It's a heuristic, not a law of physics, but it's a remarkably consistent one.

The clearest worked demonstration is Rahul Vohra's account of building the email client Superhuman, written up by First Round Review. Vohra's team surveyed engaged users (people who'd used the product at least twice in the past two weeks), and found you start getting directionally useful results at around 40 respondents, far fewer than most people assume. Their first score was an underwhelming 22%. Rather than treat that as a verdict, they segmented: they looked only at the users who were already "very disappointed," figured out who those people were, and rebuilt the roadmap around deepening the product for them while ignoring the lukewarm middle. The score climbed to 33%, then to 58% over the following three quarters.

So run the survey this quarter, then act on the "very disappointed" segment specifically. Don't average your way to a number and despair. Find the people who'd genuinely grieve the product, learn what they have in common, and build for more people like them. Fit is usually found by narrowing, not broadening.

An honest limitation. None of this is settled science. The 40% threshold comes from practitioner pattern-matching, not peer-reviewed study, and it travels badly: it suits frequently-used, self-serve software far better than enterprise tools bought by a committee or products used once a year. A survey of current users also can't see the people who tried you and left, so a high score on a tiny, self-selected base can flatter a product that's quietly failing everyone else. Use the number as an instrument, not an oracle, and pair it with retention curves and real referral behaviour, which are harder to fool.

A worked example

The figures below are illustrative, drawn to show the mechanics rather than report a real company.

Picture a four-person team with a scheduling tool for clinics. Eighteen months in, they have 600 sign-ups, polite reviews, and revenue that won't grow without discounting. The founders are days from hiring two salespeople to "scale." Instead they run the Sean Ellis survey on the 140 users who logged in twice in the last fortnight. Sixty reply. Only 24% say "very disappointed." Below the line.

The instinct is to read that as failure and add features. They resist, and segment instead. Among the "very disappointed" minority, a pattern jumps out: nearly all run single-practitioner clinics, where the owner is also the receptionist. They love that the tool collapses booking, reminders and no-show chasing into one screen they can run between patients. The "not disappointed" group is mostly larger clinics that already have front-desk staff and barely notice the saving.

That's the value hypothesis arriving in focus: solo practitioners who are their own front desk will be very disappointed to lose this, because no one else gives them back that time. The team rewrites the product story around that person, drops three half-finished features aimed at big clinics, and ships two that only a solo owner would care about. They re-survey a quarter later: 46%. Now the growth hypothesis is worth funding, and they know exactly who the salespeople should call.

flowchart LR
    A(["Survey engaged users:
'very disappointed?'"]) --> B(["Score below 40%
not fit yet"])
    B --> C(["Segment the
'very disappointed' users"])
    C --> D(["Find what they
share: solo clinics"])
    D --> E(["Rebuild for that
segment, drop the rest"])
    E --> F(["Re-survey:
score crosses 40%"])
    F --> G(["Now fund growth"])
Narrowing to the segment that already loves you is how a sub-40% product finds fit. Leaders Loop

Frequently asked questions

Is product-market fit a one-time thing you achieve and keep?

No. Fit is relative to a market, and markets move, new competitors, shifting needs, your own expansion into a new segment. A product can have strong fit with its first audience and none with the next one it targets. Treat it as a state to keep re-checking, especially after you change who you're selling to.

Can a big company have a product-market-fit problem?

Yes, and it's easy to miss because existing scale hides it. A new product, feature line or market entry inside a large firm has to find fit from scratch, just like a startup, except the surrounding revenue can fund a "push" for years, disguising the absence of pull. The same diagnosis applies: is this specific thing being pulled, or propped up?

How is this different from validation or an MVP?

Validation and an MVP are how you test the value hypothesis; product-market fit is the result when that hypothesis is proven at enough scale that the market starts pulling. An MVP can pass a usability test and still have no fit, people can find a thing usable and not actually need it. (See discovery and validation and MVP and iterative delivery for the testing craft.)

Should I add features to reach 40%, or is that the trap?

Usually the trap. Sub-40% scores are more often a targeting problem than a feature gap, you're serving a blurry "everyone" instead of the people who'd grieve the loss. Segment first. Adding features for the lukewarm middle tends to bloat the product without moving the number; building for the segment that already loves you tends to move it a lot.

What if I can't survey users yet because I have too few?

Then you're earlier than the survey is built for, and qualitative signal is your instrument: are early users coming back unprompted, and are they referring others? The survey gets directionally useful at around 40 engaged respondents. Below that, watch behaviour, real retention and real word-of-mouth, rather than a number.

Related in the Toolkit

Where to go next