Product-Market Fit: How to Know You Have It (And What to Do Before You Do)

Product-market fit (PMF) is the single most discussed and least defined concept in startups. Founders chase it like a destination; in reality, it shows up as a few specific signals, almost always at the same time. This guide names them — and explains what to focus on before they appear.

What product-market fit actually is

Marc Andreessen's definition is still the cleanest: "Being in a good market with a product that can satisfy that market." You feel it when usage outpaces your ability to keep up — servers melt, support is buried, hiring lags demand. You don't have it when growth feels like pushing a boulder uphill, no matter how good the product looks.

The 5 signals of PMF

No single metric proves PMF. The cluster does:

The 40% test. Ask active users: "How would you feel if you could no longer use this?" If 40%+ say "very disappointed", you have PMF. Below 25%, you don't. Source: Sean Ellis, validated across hundreds of startups.
Organic pull. Customers find you without paid acquisition. Word of mouth, inbound, search.
Retention curves that flatten. Day-30 or week-12 retention stops decaying. Some cohort stays forever. That cohort is your PMF segment.
NPS above 40 in your target segment (not overall).
Sales cycle compresses. Buyers stop debating. They want to know how fast you can onboard them.

If you have one of these, you have a strong signal. If you have three, you have PMF.

How to actually run the 40% test

The threshold is the easy part. Most founders get a misleading number because they survey the wrong people, the wrong way.

Who to survey. Sean Ellis's own guidance is to survey recently-active users — people who've experienced the product, not the dabblers who signed up and bounced. A practical filter: users who used the core feature at least twice in the last week or two. Surveying your whole signup list dilutes the signal with people who never felt the value, and you'll under-read.
How many. You want roughly n ≥ 30–40 qualified responses before the percentage means much. Below that you're reading noise, and one or two answers can swing you across the 40% line.
Mine the open text, not just the percentage. The number tells you whether you have fit. The free-text answers tell you who has it and why. Read the "very disappointed" cohort's open responses for two things: the pattern in who these people are (role, company size, use case — that's your PMF segment hiding in plain sight), and what they'd substitute if you vanished. If the best substitute they can name is a spreadsheet or "nothing," your moat is real. If it's a competitor a click away, it's thinner than the percentage suggests.

In the workspace, the disappointment section instruments exactly this survey, retention tracks the curve shapes below, and word-of-mouth captures the organic-pull signal — so the three hardest-to-fake signals live in one place.

Reading retention curves (the shape matters more than the number)

A single retention number is nearly useless. The shape of the curve is the signal:

A "smile" or L-curve that flattens to a non-zero asymptote means a cohort has genuinely stuck. The line stops dropping and goes horizontal — those users churned out the casuals and kept the believers. This is what PMF looks like on a chart.
Decay to zero — a curve that keeps sloping down with no flattening — means no one is forming a lasting habit, no matter how good the launch-week numbers were.

The healthy asymptote differs by category, and comparing yours to the wrong benchmark will lie to you both ways:

Consumer social / casual apps can be healthy at a lower M1 retention — a small, dense, highly-engaged core is enough if it grows virally.
Vertical SaaS / B2B expects high logo retention — losing accounts month over month is a fit problem, not a casual-usage quirk. A 70% consumer-app retention that would be excellent for social is a slow death for a $1,000/seat tool.

Don't ask "is my retention good?" Ask "does my curve flatten, and is the asymptote healthy for my category?"

Compute every signal inside the segment, not across the aggregate

This is the contradiction founders trip on: "three signals = PMF," but "PMF is segment-specific." Both are true — if you measure within the segment. Your aggregate numbers can look mediocre while one segment is screaming PMF, because the dead weight of bad-fit users drags every average down.

The order of operations matters:

Find the cohort with flat retention first. Slice your users by the obvious dimensions (role, company size, source channel, use case) and look for the slice whose curve flattens. That's your candidate segment.
Then run the 40% survey on only that cohort. Not your whole list — the people in the flat-retention slice.
Then check organic pull and sales-cycle compression for that same group.

Run the signals against the aggregate and you'll conclude you have no fit when you actually have intense fit with 15% of your users. The job isn't to lift the average — it's to find the segment that already screams, and go all-in on it.

What PMF is NOT

A press launch.
A funding round.
A viral tweet.
A waitlist of 10,000 emails (vanity).
Logo customers (they tell you nothing about whether usage sticks).

False-positive PMF: when the signals lie

PMF can be faked — usually by something that won't scale propping up the numbers. Watch for:

One non-scaling channel. A founder personally onboarding every user, a single subreddit, a friend's newsletter. Growth looks organic because you are the distribution. It dies the moment you try to systematize it.
A single enterprise whale. One big logo using you heavily makes retention and revenue look beautiful. Strip them out and the curve collapses. One account is not a segment.
Paid acquisition masking weak organic pull. If you turn off ad spend and growth stops, you've bought usage, not earned it. The 40% survey can still read high among the people you paid to acquire — which is why organic pull is a separate, non-substitutable signal.

The tell for all three: the signals depend on something you can't repeat at scale. Before you step on the gas, ask what breaks if you remove the founder, the whale, or the ad budget.

PMF decays — and it warns you first

PMF is not a one-time achievement you bank. It erodes, usually for one of two reasons: the easy segment saturates (you've signed everyone who was desperate, and the next cohort is lukewarm), or a competitor resets expectations (what felt magic now feels table-stakes).

Decay shows up in leading indicators before it hits topline revenue:

CAC creeping up — you're paying more for each new customer because the easy demand is gone.
Win-rate softening — deals you'd have closed six months ago now stall or go to a competitor.
The retention asymptote dropping in newer cohorts — old cohorts still look great (survivorship), but each new cohort flattens at a lower line. Always read retention by cohort start date, not blended, or this hides for quarters.

If your blended numbers look fine but newer cohorts are quietly worse, you're in early decay. That's the moment to narrow or re-fit — not after revenue rolls over.

The three stages before PMF

Stage 1: Problem-solution fit

You've validated that a real segment has a real problem they'll pay to solve, but you haven't built anything sticky yet. Focus: customer interviews, demand tests, manual delivery of the solution.

Stage 2: First sticky users

You have 5–20 customers who actively use the product and would notice if you took it away. Don't scale yet — this is where most founders break PMF by chasing growth before the product is sticky. Focus: deepen usage, find the "magic moment" in onboarding, kill features the sticky users don't touch.

Stage 3: The PMF threshold

The signals above start appearing. Now you can step on growth — but only on the segment showing the signals. The classic mistake here is generalizing too fast. PMF is almost always with a narrow segment first, not your full TAM.

What to actually do at each stage

Pre-PMF. Talk to users every week. Personally. Read every support ticket. Cut features. Don't hire a head of sales.
At PMF threshold. Pick one growth channel and double it. Hire to remove the bottleneck (often you).
Post-PMF. Build the system — onboarding, success, sales — that lets non-founders deliver what you've been delivering by hand.

The hardest part: knowing when to stop iterating

Founders pre-PMF often pivot too late — they keep building features hoping retention curves bend on their own. They don't. If you've spent 6+ months and retention is still decaying, the product or the segment is wrong. Either change the segment (same product, different customer) or change the product (same customer, different solution). Don't change both at once.

PMF is segment-specific

You don't have "product-market fit". You have fit with a segment. The segment that gives you PMF is usually narrower than your original target — and that's a good thing. Document who that segment is, and write everything (positioning, pricing, sales scripts) for them, not the broader market you wish you served.

The decision rule: narrow, re-fit, or change the product

When the signals are ambiguous, don't agonize — read the two numbers that disambiguate: retention by cohort and "very disappointed" by sub-segment. They point to different actions:

Decaying retention overall, but a sub-segment with high "very disappointed" and flat retention → NARROW, don't pivot. You have real fit hiding inside a too-broad target. Cut the target to that segment and rewrite positioning, pricing, and sales for them. This is the most common situation and the most commonly misread as "we need to pivot."
Low "very disappointed" everywhere, no segment with a flat curve → change the product. No one would truly miss you. That's a product or solution problem, not a targeting one. Same customer, different solution.
A segment screams but you can't reach more of them affordably (CAC too high) → it's a channel problem, not a fit problem. Don't touch the product; go find distribution.

The discipline from earlier still holds: change the segment or the product, never both at once — or you'll never know which move worked.

The fastest way to lose this thread is to drift. A one-page Product Vision Document — re-read at the start of every roadmap meeting — keeps you honest about who the segment is and what the kill/narrow numbers are. Use it as your operating compass until the PMF signals are unambiguous, then again every time the leading indicators wobble.