The Search for Real Product-Market Fit (PMF)

One of the sentences heard most often in early-stage teams is “we're close to PMF.” It usually arrives with evidence: the waitlist is long, a publication wrote about us, a few customers said “love it,” MRR is climbing. Then someone opens the cohort table and it tells a different story: say, by month three half the users have quietly left, and the team sees it for the first time in that meeting.

That is why I approach PMF through behavior, not belief. Fit is the market telling you, with wallets and habits, that your product fills a gap. It starts when they say it unprompted, not when you say it for them.

When you need it

Finding fit is live work from the moment the product meets its first users, but there are concrete signals that you should switch into deliberate search mode.

• You have growth, but the bucket leaks: new signups keep coming while active users hit a ceiling.
• In sales calls, every customer says yes for a different reason and you cannot extract one shared story.
• The roadmap is being shaped by whoever asks loudest.
• When you raise marketing spend, quality drops rather than volume rising.

The sign that it is still too early is simpler: you do not have enough users to analyze. Drawing a retention curve from fifteen users is meaningless. At that stage the work is not statistics, it is conversation. The goal is to get one group of users to say, “if I lose this, it hurts.”

What most teams miss

PMF is not a binary state, it is a property of a segment. “Has our product reached PMF?” is the wrong question. The right one is “in which group of users, under which conditions, has it?” There is no average user. A product that shows no fit on average can hide a very strong fit in one segment, and teams miss it precisely because they look at the average.

The approach Superhuman's founder Rahul Vohra has described is a good illustration. They used Sean Ellis's question: “How would you feel if you could no longer use this product?” The commonly cited threshold is that roughly 40% should answer “very disappointed.” But the real move was not reading the score, it was splitting it. They segmented respondents, studied what the “very disappointed” group loved, and pointed the roadmap at strengthening that, not at winning over the fence-sitters.

Do not stop at measuring PMF. Find out who has it, then go after everyone who looks like them.

The second misunderstanding is about time. Teams treat fit as a destination: one day we say “we're there” and shift into growth mode. But fit erodes as the market and your competitors move. When a new entrant solves the same problem at half the price, or your customer's way of working changes, last year's flat curve starts bending again this year. The teams that open the cohort table every month, not once, notice before they lose it.

A distinction here. This piece is about the evidence for fit and the discipline of searching for it. Pinning down exactly what the product solves, for whom, under which conditions, is a separate job, covered in the product-value fit piece. Evidence first, definition second.

The playbook

1. Define what proof looks like

Before you start, write down what result would make you say “fit exists.” For example: an eight-week cohort retention curve that flattens instead of reaching zero, a meaningful share of new users arriving by word of mouth, and the Ellis survey clearing its threshold. Writing thresholds before the data, not after, makes it harder to fool yourself.

2. Slice cohorts by segment

Break weekly signup cohorts by channel, company size, use case and first action taken. You are looking for any slice whose curve flattens. If even one does, fit has started to sprout there.

3. Talk to the people who feel it

Interview at least fifteen to twenty active users: how they first heard of you, which alternative they dropped, what they would do without the product, who they told. The answer to “what would you do without it” is the honest measure of your value. “I'd go back to Excel” and “I'd hire someone to do this by hand” are not the same answer.

4. Cut the roadmap around the fit group

Analyze the Ellis survey's “main benefit?” question only among the “very disappointed” respondents. Deliberately decline most of what the fence-sitters ask for. This is uncomfortable because it means saying no to part of the demand. Teams that skip this step can find themselves, six months later, with a product that serves everyone a little and that nobody would miss.

5. Lock growth onto the fit segment

Point marketing and sales only at the profile with proof of fit. Chasing volume before fit is clear means pouring a full bucket into a leaky one.

What to measure

• Floor of the cohort retention curve: Does it reach zero or flatten? A lagging but the most honest indicator.
• Ellis score, by segment: The share answering “very disappointed.” A leading indicator, readable within weeks.
• Share of organic and referred signups: Is demand arriving on its own, or are you buying it?
• Time to activation: How many users reach first value, and how fast. The earliest trace of fit inside the product.

Pitfalls

Looking at the average. If overall retention is 20% and one segment retains at 60%, your product has two stories. The average hides it.

Trusting vanity metrics. Signups, traffic, press, waitlist. None of them tells you whether users come back. Celebrate them, but do not make decisions with them.

Purchased fit. A customer base held together by discounts, heavy sales support or hand-carrying is not fit. Ask what remains when the support and the discounts go. That is the exam.

Start this week

1. Split the last six months of signups into monthly cohorts and write each cohort's week 1, week 4 and week 8 active rate on one page.
2. Send the Ellis survey to people who used the product at least twice in the last 30 days, and add one question: “How would you describe this to someone else?”
3. Book twenty-minute calls with your five heaviest customers and ask all of them the same thing: “What would you do if this product did not exist?”