Product-market fit meaning: a plain-English definition for founders (+ free cohort retention chart builder)
- BADideas.fund
- 3 days ago
- 9 min read
Updated: 3 days ago
The product-market fit meaning most founders get wrong: PMF is a retention curve, not a feeling. How to read yours, and why a curve that bleeds to zero is often a go-to-market problem rather than a product one.
I'm Jurģis - community & brand lead at BADideas.fund. This issue kicks off a new format for the newsletter: each week we take one bit of startup jargon and gut it: what it means, what everyone gets wrong, and what to do about it.
The short version, for the people in a hurry. Product-market fit is when a group of customers keeps coming back without you pushing them. It shows up as cohort retention that flattens, not as a good launch week. Take everyone who signed up in the same week, track what share are still active at weeks 1, 2 and 6, and read the shape. If the curve settles onto a plateau, even a low one, you have fit with somebody. Google Photos flattened between 20% and 40% depending on country and device, and that was enough to build on. If the curve slides toward zero, you do not have fit, and adding signups will hide that for roughly two quarters. The second half of the answer matters more: a bleeding curve is not automatically a verdict on your product. Often you are acquiring the wrong users, and that has a completely different fix. There is a free tool further down this page that builds the chart and the segment split from your own data, in your browser.
Every founder I meet has early product-market fit. Very few can tell me what would make them admit they do not. That is what the phrase has turned into: a status you award yourself rather than a bar you clear. So here is the bar, and then the question that actually decides what you do today.
Product-market fit meaning: the one-line definition
Product-market fit is when the market pulls the product out of your hands. People adopt it, keep using it, and tell other people, without you in the room. The cleanest working test I have heard comes from Jason Cohen, who founded Smart Bear and WP Engine. When he diagnoses a company that has stopped growing, he does not open with revenue or pipeline. He opens with whether customers stay:
"Step one is this logo retention, essentially do we have product market fit?"
Retention first. Then pricing and positioning. Then growth. His point is the sequence: run those three in any other order and you are optimising a leak. Spending on acquisition before the retention question is answered makes the top-line number rise while the product quietly fails everyone who arrived last month.
Product-market fit is a curve, not a moment
Fit is not a line you cross once and announce. It is a shape you watch. Group users by the week they arrived, then chart what share are still active 1, 2 and 6 weeks later. You are looking for the curve to stop falling. A curve that flattens means a real group of people found lasting value and stayed. A curve that decays to zero means you are renting attention. The Y Combinator Startup School session on retention puts the test in one sentence:
"If your curves don't flatten out, I would say it's a pretty good sign that you haven't yet made something people want."
Two details decide whether your chart tells you the truth. First, the window has to match how the product is meant to be used: daily for high-frequency products, weekly or monthly for a utility someone opens when a specific job appears. Measure a monthly tool on daily retention and a healthy product looks dead. Second, you need to know which single action predicts that someone comes back. Sarah Tavel, a partner at Benchmark who was early at Pinterest, calls it the core action:
"It's an action that if they perform the action they're very likely to come back."
For Facebook that action was friending. For Pinterest, pinning. Before fit, most of your work is finding your version of that action and getting more new users to reach it, sooner.
The misdiagnosis that costs a year
Here is where most founders lose time, and it is the part the definition never covers. A curve that bleeds to zero gets read as a product verdict. The founder concludes the product is not good enough, and goes back to building. Six months of roadmap later, the curve looks the same.
Sometimes the product genuinely is the problem. Often it is not. The same Y Combinator session names the other cause directly:
"You've built a great product but you're targeting it to the wrong type of customer."
If you are pouring the wrong people into a product that works for a narrower group, the aggregate curve will bleed no matter how good the software gets. The fix is not in the codebase. It is in who you acquire and what you say to them, which is a go-to-market problem with go-to-market levers: the ICP definition, the message, the channel, the sequence in which you test them.
This is the belief the whole BADideas thesis sits on. At seed, great founders with real direction still die, and they rarely die on the product or on what they believe. They die on go-to-market: on which experiments they run, in what order, and how well they run them. Moving from your first handful of customers to a repeatable way to acquire hundreds more is the entire game between now and your next round. Get the bets wrong about what that engine is and it does not slow the company down, it ends it before the product ever gets a real shot.
The practical version, before you touch the roadmap: split the curve. Chart retention separately by segment, by acquisition channel, and by how the user arrived. If one segment flattens at 35% while the blended average bleeds, you do not have a product problem. You have a targeting problem, and you have just found the customer you should be building the company around. One founder in our portfolio, Leszek at Juo, had demos converting well and revenue refusing to follow. The product was not the issue. The ICP had been drawn too wide, so the company was selling to anyone who might plausibly benefit rather than the profile that converted consistently. Redefining that ICP, rebuilding the messaging around it, and sequencing the path from interest to revenue is what moved the numbers.
Why a bleeding curve stays hidden for three months
There is a second reason founders misread the curve, and it has nothing to do with analytics. Naming a retention problem out loud has a cost, especially to the person who wrote you a cheque.
On a Tuesday morning a founder drafts three sentences to his investor about why the ICP is not working and conversion is down. He reads them back, decides the investor will not know what to do with this, deletes them, and sends something manageable instead. The real situation does not get named until it is three months old and considerably harder to fix. At seed, three months is material. Martin, who runs Fleetfox, described the dynamic without being asked:
"With other VCs I feel like I have to draft a message and make sure I don't say anything wrong."
That instinct is rational when your investor asked for a five-year projection on a four-month-old product. It is also the mechanism by which a fixable retention problem becomes a fatal one. A founder performing strength gives everyone a managed version of the problem, and a managed problem gets a managed diagnosis. If you are reading your own curve this week, the useful question is not only what the chart says. It is who you can show it to on Tuesday instead of in October.
How to measure it honestly
Three practices, drawn from what consistently shows up across Cohen, Tavel and the Startup School material.
Start with cohort retention, not revenue. One chart: rows are signup week, columns are weeks since signup, cells are the share still active. It answers the fit question more honestly than anything else on your dashboard, and the tool below will build it from whatever export you already have.
Name your comeback action and instrument it. Write down the single behaviour that predicts a return visit, measure what share of new users reach it, and treat "more new users to that action, sooner" as the job until the curve flattens.
Segment before you conclude. Never act on a blended curve. Split by channel and customer type first, because the blended average is what turns a targeting problem into a rebuild.
Underneath all three is the thing that actually compounds: how many real hypotheses you can test in a week, and how fast you update when the market answers. Not motion for its own sake. Learning rate. Founders close to the problem do not just run more experiments, they run better ones, because they know which hypotheses are worth testing at all.
The tool of the week: a cohort retention chart builder
The third of those practices, segmenting before you conclude, is the one founders skip. Not because it is conceptually hard, but because it means wrestling a spreadsheet into a triangle and then doing it again for every segment. So we built it instead.
It needs two columns: a user id and a date. Activity means whatever counts as using your product, so a login, a session, an order, an invoice or an API call all work. If you have a signup date, map it and the tool uses it. If you do not, it takes each user's first activity as their cohort. Add an optional third column, a segment such as plan, channel, country or industry, and it does the part this article is actually about: it draws the blended curve, splits it by segment, and tells you when the two disagree. On the sample data it reports that self-serve holds at 35% while outbound falls to 2%, and that the blended 17% describes neither of them.
Two things worth knowing before you paste anything in. It runs entirely in your browser, so your customer data is never uploaded and the page works offline once loaded. And it excludes cohorts that have not lived long enough to reach a given week rather than counting them as zero, which is the most common way a hand-built retention chart flatters you.
What to do this week
Build one cohort retention chart, using the tool above or a spreadsheet, and look only at whether week 6 is flat or falling. Ignore the absolute number for now.
Split that chart by acquisition channel and by customer type. If any single segment flattens, stop planning product work and go read that segment.
Write down your comeback action and the share of new users who reach it. If you cannot name the action, that is this week's work.
Send the Tuesday email. Whatever the curve says, tell the person who funded you now rather than in three months.
Frequently asked questions
What does product-market fit (PMF) mean?
Product-market fit means your product satisfies real demand strongly enough that a group of customers keeps using it without being pushed. In practice it shows up as cohort retention that flattens onto a plateau rather than decaying toward zero.
How do you measure product-market fit?
Group users by the week they joined and track what share stay active over the following weeks. If the curve flattens at a stable level, that is evidence of fit. If it keeps falling to zero, you do not have it yet. Retention is a more reliable signal than signups or revenue, both of which can rise while the product is failing new users.
Is product-market fit a feeling or a metric?
It is a metric. Excitement on sales calls, a rush of demo requests and a strong launch week can all happen without fit. Fit is repeated usage visible in the data, behaviour that continues when the founder is not pushing.
What's the difference between product-market fit and idea validation?
Idea validation happens before you build: cheap conversations to decide whether an idea is worth pursuing at all. Product-market fit is measured after people use the product, and it asks whether they keep coming back. You validate to decide what to build, then read retention to find out whether it fits.
How much retention means you have product-market fit?
There is no universal number, because it depends on how often your product is naturally used. The signal is the shape, not the percentage. Google Photos flattened between 20% and 40% by country and device, which was enough. A curve that keeps declining is not fit at any percentage.
My retention is bad. Is my product wrong?
Not necessarily. A bleeding curve has two common causes: the product does not deliver lasting value, or you are acquiring the wrong customers for a product that works well for a narrower group. Segment the curve by channel and customer type before you conclude. If one segment flattens while the average bleeds, the problem is targeting, and rebuilding the product will not fix it.
Work with BADideas.fund
We're an early-stage fund backing founders across CEE who are betting on things that sound wrong right up until they are obviously right. We are ex-founders, and our thesis is the one running through this piece: seed companies rarely die on the product or on what the founder believes, they die on go-to-market, on which experiments they run, in what order, and how well they run them. So that is where we go all in. Within weeks of closing, every company runs a structured GTM diagnostic, and when it surfaces a specific problem, a named operator who has run that exact function before, in outbound, growth or product-led, is briefed and in the room. You bring the read on the customer. We work on the translation from that read into the right experiment, in the right order, faster than you would get to it alone.