Product-market fit: definition, signals and method

Product-market fit: the definition

Having found market fit means having built a product that delivers value to a specific group of people, who adopt it and are willing to pay for it, in a repeatable way. The full term is product-market fit: the match between a product and a market, popularized by Marc Andreessen in 2007. The market pulls the product.

It is the first stage in a product’s life, and a product manager’s core mission for as long as it has not been reached. Before market fit, everything else is secondary. After it, everything changes: the challenge becomes growth and optimization.

What market fit is not

• Enthusiasm in a demo. People are polite, and a good demo impresses. It says nothing about usage the following Tuesday.
• A funding round. Investors have bet on your ability to find it, not observed that you have found it.
• A few big customers. Two contracts signed through your network prove your network. Market fit means the sale repeats without you.

The 4Ps of the marketing mix

For a product aimed at a broad market, fit covers the 4Ps of the marketing mix:

1. Product: what it does, its value proposition.
2. Price: what the target is willing to pay.
3. Place: how the target gets access to the product.
4. Promotion: how the target hears about it.

At the start, you do not need all four. Use your MVP to validate the first two: product and price. Only later, once you know what the product has to be for the target to buy it, does investing in distribution and promotion pay off. That is when sales and marketing teams come in. Not before.

How do you know you have found it?

The symptoms

The usual metrics give you clues: a user base that grows without proportional effort, regular engagement, a retention curve that flattens out instead of sliding to zero, revenue that rises steadily. The Sean Ellis test asks users how disappointed they would be if they could no longer use the product (the threshold is 40% “very disappointed”).

These signals are useful, but they are symptoms. You can produce them temporarily by buying growth.

The most reliable symptom remains the cohort retention curve. Group users by the month they arrived and track, for each group, the share still active in the following months. If the curves drop and then level off on a plateau, a core of users has found a reason to stay. If they trend towards zero, no amount of acquisition will make up for the leak.

The fundamentals

I prefer three tougher criteria, reasoned from economic fundamentals:

• You make money on every sale, once direct costs are counted.
• More customers means more margin: your marginal cost stays flat or falls.
• Your way of acquiring customers is repeatable: you know what each euro invested brings back.

If any one of the three is missing, you do not have market fit yet. You may have enthusiastic customers, which is excellent news, but it is not the same thing.

A worked example with numbers

A fictional example: inventory management software for independent bakeries, sold at €90 a month.

1. Margin per customer: hosting, support and payment fees cost about €25 per customer per month. That leaves €65. First criterion met.
2. Marginal cost: support makes up most of the direct costs. If every new customer generates as many requests as the first ones, the margin stalls; if documentation and onboarding reduce those requests, it grows. The team tracks the number of requests per customer per month.
3. Repeatable acquisition: the first twenty customers come from word of mouth. A test with a bakery equipment distributor brings in fifteen more at about €400 each, paid back in just over six months of margin. It is this channel, measured and repeatable, that moves the company from “happy customers” to “market fit”.

The numbers are made up, the logic is not: three questions, three measurements, and an answer that can be challenged with data.

A checklist

• You can describe your typical customer in one sentence, and your best customers recognize themselves in it.
• Cohort retention curves level off.
• You know your margin per customer, direct costs included.
• At least one acquisition channel works without the founders stepping in.
• The customers you lose leave for reasons you understand.

How to find it

Many products never get there because the team focuses on what it wants to build, not on what the market needs. The classic symptom is the feature race: “we’re missing this feature, so of course we have no customers.” That reasoning almost never holds up.

Win a niche first

The most useful advice comes from Geoffrey Moore (Crossing the Chasm, 1991): focus on the needs of one specific niche. Who are you building for? What does that group need? Talk to them, have them try your MVP, gather feedback, and build to dominate that niche before going after another one.

Work through hypotheses

The second piece of advice: write down the list of hypotheses that stand between you and market fit, and how you will validate each one. That is the hypothesis roadmap. It stops you from building gadgets that complicate the code and blur the value proposition, and in hard times it tells you whether you are stuck or making progress that does not show yet.

Measuring your progress

Measure two things side by side. On one hand, hypothesis validation: if none has been validated or invalidated in a month, you are stuck. On the other, product metrics: activation, retention, conversion, churn. As long as both are moving, you are making headway, even with setbacks.

Three challenges always come back: not getting distracted by the search for the perfect solution, persevering when results are not yet visible, and accepting a pivot when an important hypothesis falls. A pivot is not necessarily a change of business model: adjusting the price or the niche is often enough.

Typical mistakes, and how to spot them

1. Mistaking bought growth for traction. Sign: sign-ups rise with the ad budget and fall back when it stops. Look at cohort retention, not at the volume of arrivals.
2. Averaging different segments. Sign: retention that looks “decent” on average. Break it down by segment: often, one segment loves the product and the others leave. Market fit lies in the first.
3. Forgetting the cost to serve. Sign: revenue grows, cash does not. Count the support, onboarding and customization time per customer.
4. Declaring victory too early. Sign: the team hires salespeople while sales still depend on the founders. Wait until someone from outside has sold using the same method.

Small team or large organization

In a startup, the search for market fit is a matter of survival, and discipline comes naturally: cash sets the pace. The risk is rather pivoting too often, without giving a hypothesis time to be tested.

In a large organization launching a new product, the danger is the opposite. The internal sponsor, existing customers and the sales force can create artificial signals: customers who sign out of loyalty, internal users who have no choice. Apply the same criteria while isolating these effects: measure usage among customers who were not guided by their usual account manager, and adoption among teams that could have said no.

After market fit

Once you have found fit with your first users, a new phase begins: winning a broader market with different expectations. That market wants reliability, compatibility and clear pricing, not novelty. Product work changes in nature: fewer features, more stability, better service quality and metric optimization. The product roadmap changes its purpose accordingly.

The classic mistake is to believe that the product that won over early adopters will win over the mainstream market as is, and to overinvest in growth too early. Keep a cool head.

Where does AI fit in?

AI changes two things in the maths. First, it makes demos very convincing: an AI product can generate instant enthusiasm in a demo without ever becoming a habit. The only signal that counts is repeat usage, in real conditions, by people who were not involved in the project.

Second, it has a real marginal cost: every request costs money, and that cost varies with usage. The second criterion above, a margin that holds as volume grows, becomes a design question, not just a finance one. I check this systematically before any production launch: see from AI POC to production and measuring an AI assistant.

In hindsight

I would keep the 2022 definition, and above all the preference for fundamentals over symptoms. It is the passage of this article I quote most often. Spectacular growth can hide a negative margin; modest growth can rest on sound foundations.

I went through a period of hyper-growth at TIAO, a B2B marketplace whose user base grew twelvefold in two years. That kind of period makes the distinction even more useful: when everything is going up, it becomes hard to know what is really pulling, and easy to credit the product with what comes from the market or from the sales effort.

I would nuance the order I proposed: product and price first, distribution later. It still holds, but today I would look at distribution earlier. At SipScience, I worked on the roadmap, but also on acquisition channels and the onboarding journey. A sound product with no repeatable path to its customers does not have market fit yet, and finding that out late is expensive.

I see what AI assistants have changed from both sides. On the method side, testing a value proposition costs less: a page, a prototype, a series of interviews synthesized in a few hours. On the product side, the demo trap has got worse. An AI product can look remarkable on ten examples and disappoint on the hundredth real case. So you need to check the outputs, measure repeat usage, and track the cost per request from the first customers onwards. And, as with any AI transformation, remember that adoption depends more on the organization than on the technology: 70% versus 30%, in my experience. A B2B customer adopts an AI tool when its teams have changed the way they work, not when the model has improved.

For an internal tool, the same logic applies with a simpler question: is the time saved real? The guide to measuring the time saved by an AI tool sets out the method.

Frequently asked questions

What is the difference between market fit and product-market fit?

In practice, none: “market fit” is the short form of “product-market fit”, the match between a product and its market.

How long does it take to find market fit?

There is no typical timeframe. What matters is how fast you validate hypotheses: a team that tests one a month learns far faster than a team shipping features with no hypothesis behind them.

Can you lose market fit?

Yes. A shifting market, a competitor, or a technology such as generative AI can make a value proposition obsolete. Retention and customer acquisition cost are usually the first metrics to show it.

Is the Sean Ellis test enough to prove market fit?

No. It is a good indicator of attachment, but it measures an opinion. Complement it with behavior: cohort retention, margin per customer and repeatable acquisition.

How do you assess market fit for an AI product?

As for any product, with two points of attention: repeat usage by people outside the project, on real cases, and the cost per request, which must leave a margin as volume grows.