Product Market Fit: How to Find It and Know You Have It

An MVP has been live for eight weeks. It has 180 sign-ups, two enthusiastic emails and a founder who has started saying the team is close to product market fit.

Nobody has written down what close means.

Most advice on product market fit doesn't help at this stage. It describes the state after the fact: customers buying as fast as you can build, reporters calling, hiring at full speed. By then, nobody needs a test.

The useful question for a young product is smaller: whether the people who use it would miss it, and whether that group is growing. You can measure that with a few dozen responses. Most teams don't, because the usual instruments assume thousands of users.

This article is for the founder or product lead with an MVP in market and a few hundred users. It replaces "it feels like it" with something you can measure.

Key Takeaways

• Product market fit means a defined group of customers would be upset to lose your product and keeps using it without being prompted.
• The Sean Ellis survey asks active users how they'd feel if they could no longer use the product. The usual benchmark is 40% answering "very disappointed".
• You don't need thousands of users to run it. Superhuman's Rahul Vohra reported directionally useful results from around 40 respondents.
• Retention and unprompted referrals back up the survey. Sign-ups, feedback from friends and revenue from one or two customers can all mislead.
• A low score isn't a verdict. Find who answers "very disappointed", learn why, and build for that group first.

What Product Market Fit Means

Product market fit is when a defined group of customers needs your product enough to keep using it. They'd be upset to lose it. And they tell other people about it.

Marc Andreessen's 2007 essay gave the idea its most quoted definition: a good market, plus a product that can serve it. The market comes first in that definition.

He also described how fit feels. Customers buy as fast as you can build. Usage grows as fast as you can add servers. The absence looks different: customers don't quite get value, and word of mouth doesn't spread.

Those are lagging signals. By the time they show up, you already have fit. A team with a young MVP needs something earlier. That's where measurement comes in.

Two points about the definition matter for what follows. Fit is tied to a specific market, so a product can have it with one segment and not another. And it isn't a switch. It builds over time. But it can slip when you reach beyond early adopters.

A minimum viable product exists to put working software in front of real users. That's what makes any of this measurable. Product validation comes earlier and tests whether an idea deserves building. Fit is about what happens once users have the product.

How to Know if You Have Product Market Fit

You have product market fit when a large share of your regular users would be upset to lose the product. Two signals, read together, get you most of the way there. One is what users say. The other is what they do.

The Sean Ellis Survey

Sean Ellis ran early growth at Dropbox, LogMeIn and Eventbrite. He found a single survey question that separates products with traction from products without it. The question asks how users would feel if they could no longer use the product. The options are very disappointed, somewhat disappointed and not disappointed.

Rahul Vohra's First Round essay on Superhuman describes what Ellis found. After benchmarking nearly a hundred startups, he saw that companies struggling to grow almost always had fewer than 40% of users answering very disappointed. Companies with strong traction almost always cleared it.

Slack is a useful reference. Hiten Shah put the question to 731 Slack users in 2015. 51% said they'd be very disappointed to lose it.

Clearing 40% is harder than it sounds.

And who you survey matters as much as the number. Ellis recommends people who've used the product at least twice in the last two weeks. They've experienced the core of it. A sign-up who never came back can't say whether they'd miss it.

What a Small Sample Can Tell You

Superhuman polled between 100 and 200 users when it first ran the survey. Vohra writes that results start to look directionally correct at around 40 respondents, which is fewer than most teams assume.

But directional is the right word. At 40 respondents, the margin of error on a percentage is roughly 15 points either way at 95% confidence. A result of 15% and a result of 55% tell different stories. A result of 38% and one of 42% tell the same one.

Ellis also cautioned against leaning on surveys too early. In his 2012 blog post, he wrote that young startups often lack enough users to automate feedback. He added that interviews should keep complementing surveys even after that. So run both, and treat the score as one input.

Retention and Referrals

What users say is one half. What they do is the other, and it's harder to fake.

Group users by the week they joined and track how many are still active in later weeks. A curve that flattens means a core group has made the product part of their routine. A curve that keeps sliding toward zero means nobody has. Aggregate active-user counts hide the difference, because new sign-ups replace the people who left.

Cohorts can't be rebuilt from data you never collected. Add the analytics before launch, while you're deciding how to build an MVP, not after.

Unprompted referrals are the second behavioural signal. When users tell other people about the product without being asked, that's the word of mouth Andreessen described. It's slow to appear. But it's difficult to manufacture. In our experience, it usually shows up before revenue does.

Signals That Look Like Fit but Aren't

Several things feel like fit and aren't. Each has a cheap check.

Feedback from your own network is the most common. Friends, former colleagues and early supporters answer kindly and stay a little longer than strangers. Survey them separately from users who found you on their own, and treat the stranger group as the real number. It's one of the failure modes in the wider MVP development process, where it's called the friendly-audience test.

And sign-ups mislead in a different way. A landing page that converts well can front a product nobody keeps using. Count the people who complete the core action more than once.

But revenue can mislead too. If one customer accounts for most of your income, you may have found a good customer rather than a market. Check whether the next ten look like the first.

Applause for a demo works the same way. People are kind about things they don't depend on. That's why the MVP vs prototype distinction matters before you run any survey.

Finding Product Market Fit When Your Score Is Low

A low score is a starting point, not a verdict. Superhuman's was 22% when it first ran the survey. Its founder, Rahul Vohra, has described what the team did next, and the pattern is worth borrowing.

It started by segmenting. The team looked at who had answered very disappointed and what those people had in common. Narrowing the market to those personas lifted the score to 33% before any product changes. But that only changed how the score read. Raising it took product work, and the first question was why the core group loved the product. The same group's answers about the main benefit showed what to protect. For Superhuman, that was speed, focus and keyboard shortcuts.

Then came the fence-sitters. Users who said somewhat disappointed and named the same main benefit were close. Their suggestions showed what held them back. The team set aside feedback from users who said not disappointed. Building for them would have pulled the roadmap away from what the core group valued.

The roadmap then split roughly in half. One half went on what the core group loved. The other went on what held the fence-sitters back. Vohra's reasoning was simple. Doing only the first wouldn't raise the score. Doing only the second would let competitors catch up.

Within three quarters of product work, the score nearly doubled from 33% to 58%.

Two cautions from the same essay apply to any team. Early adopters are forgiving, so scores can fall as you reach users who expect more. And each person should be surveyed only once, or the results get skewed.

A Worked Example With 46 Responses

Here's how it plays out with invented numbers. An MVP has 70 active users, and 46 of them complete the survey. Thirteen say very disappointed, which is 28%. That's below the bar.

Now segment by who respondents say the product is for. Seventeen describe themselves as freelancers, and nine of them said very disappointed. That's 53%. The other 29 respondents produced four very disappointed answers, or 14%.

The overall score says no. The segment score says maybe. With 17 respondents, the freelancer figure carries a margin of error of around 24 points. So it's a lead, not a result. The next step is to find more freelancers, re-run the survey with them and see whether the number holds.

Keep the cuts coarse. With 46 responses you can afford two or three segments, not a dozen.

What Comes After Product Market Fit

Once one segment clears the bar, the question shifts from finding fit to building on it. Andreessen split a startup's life into before and after fit. And the priorities change on the other side.

The next build is usually a minimum marketable product, made for the segment you've proven. But growing before that point means paying to acquire users a product can't keep.

Let's Sum Up!

Product market fit isn't a moment you announce. It's a measurement you take, repeat and act on. And a young MVP can take it with a few dozen responses and a cohort table. It takes a willingness to hear a low number, too.

Run the survey this month, not in the quarter you feel ready.

Classic Informatics has spent 23+ years helping founders and product teams build a first version and read what it says. If you're planning an MVP built to test fit, our MVP development services put the measurement in from the start. And we're happy to talk it through.