Product-market fit isn't a vibe. It's a hypothesis you can test.

Product-market fit is the most-used phrase in startups and one of the least understood. Everyone nods along. Nobody agrees what it actually means. And the confusion isn't harmless — it's the reason teams burn months "chasing PMF" with no way to tell if they're getting warmer or colder.

The root cause is simple: there are too many definitions floating around, and most founders have absorbed a blurry average of all of them. So let's get specific. Here's what the people who coined and popularized the term actually said.

What the experts actually mean

Four definitions worth knowing:

• Andy Rachleff (who coined the concept) frames it as a value hypothesis — the key assumption about why a customer would use your product. Find a compelling value hypothesis, and you've found the features to build, the audience that'll care, and the business model that makes them buy.
• Marc Andreessen (who popularized the term): "being in a good market with a product that can satisfy that market."
• Eric Ries (The Lean Startup): the moment a startup finally finds a widespread set of customers that resonate with its product.
• Steve Blank (The Four Steps to the Epiphany): having a product in a market with real, identifiable, addressable customers.

Pick one and use it across your whole team. Mixing definitions is how you end up with the sales lead, the founder, and the head of product all claiming different things about whether you've "got it." Personally, I lean on Andreessen — because he refuses to let you separate the product from the market it lives in.

Why the market half is the part you keep ignoring

Here's the trap most product people fall into: we fall in love with the product. We polish it, we obsess over the experience, we A/B test the onboarding. Meanwhile the market — is it big enough, is the problem urgent enough, are these customers reachable and willing to pay — gets treated as a given.

It isn't a given. The harshest truth in Andreessen's framing is that you can build a genuinely excellent product and still have no fit, because the market underneath it is too small, too indifferent, or too broke. Fit is a relationship between two things. Perfecting one of them doesn't get you there.

Turn it into something you can test

This is where Rachleff's "value hypothesis" earns its keep. If PMF is a hypothesis, then you stop "chasing a feeling" and start running experiments. State it plainly:

We believe [this specific type of customer] has [this urgent problem], and will [adopt / pay for / stick with] our product because [this core value] — and we'll know we're right when we see [this behavior].

Now it's falsifiable. You have a customer to talk to, a problem to validate, and a behavior to watch for. The whole machinery of product discovery exists to test exactly this kind of statement — cheaply, before you've bet the company on it.

How to tell you're getting close

You don't cross a finish line and hear a bell. But the signals are real, and they're behavioral, not emotional:

• Customers get visibly upset at the idea of losing the product. (The classic test: would they be "very disappointed" if it disappeared?)
• Retention curves flatten instead of decaying to zero — people don't just try it, they keep coming back.
• Usage pulls ahead of your ability to keep up. Demand starts dragging the company, instead of the company pushing demand.
• Word of mouth shows up uninvited. People recommend it without you asking.

Notice none of those are "we shipped a lot of features." Fit shows up in what customers do, repeatedly, on their own.

The point

Companies often go through many iterations before they find product-market fit — if they ever do. That's not a reason to despair. It's a reason to make the target concrete, so each iteration teaches you something instead of leaving you guessing.

So before your next planning cycle, do one unglamorous thing: write down your PMF definition and your value hypothesis, in one sentence each, and get the whole team to agree on them. You can't hit a target you won't name. Name it, and every experiment after that finally points the same direction.