Your Retention Curve Is the Only PMF Signal That Matters

There is one question a retention curve answers, and it is the only product-market-fit signal that cannot be bought: does it flatten, or does it head to zero?

A curve that flattens means some group of users found something they keep coming back for. A curve that keeps falling means people try the product and leave, however good the early numbers look.

Flattening beats level

This is the part that surprises people. A curve stabilising at 25% is a better business than one starting at 70% and still falling at month six.

The first has a floor. There is a group the product genuinely serves, and every acquisition dollar adds permanently to a base. The second has no floor yet, and until it finds one, acquisition is refilling a bucket rather than growing one.

Absolute retention tells you how big the core is. Flattening tells you whether there is a core at all. The second question comes first.

The benchmarks

That last row is the flattening test. If you are still losing three or four points a month at month six, the curve has not found its floor and projecting it forward is not encouraging.

Measure usage, not contracts

Logo retention on annual contracts can look excellent right up until renewal, because a customer who stopped using the product in month two is still counted for ten more months. Usage retention is the honest early signal. If you have both, look at usage first and treat logo retention as the lagging confirmation.

The same applies to defining activation. "Created an account" is not activation. Activation is the first moment a user experienced the thing the product is for — the invoice sent, the report generated, the teammate invited. Teams that define it as login report flattering numbers and then cannot work out why paid conversion is bad.

When it is not flattening

Before concluding you lack product-market fit, segment. Blended curves routinely hide one segment retaining beautifully and another never activating at all. The blend averages them into something that looks uniformly mediocre.

Split by:

• Acquisition channel — paid social traffic and inbound referrals almost never retain alike.
• Company size — the most common place a real split hides.
• Use case — especially if your product does two things and one of them is stickier.
• Onboarding path — self-serve against hand-held.

Very often one segment flattens well above the blended line. When that happens, you do not have a retention problem. You have a targeting problem, and the fix is narrowing the ICP to the group that already works rather than making the product fit everyone.

Why this decides where money goes

The practical consequence is an ordering rule.

If the curve is heading to zero, acquisition spend leaks back out. More traffic produces a bigger number this month and the same number in six. Retention has to come first, and that is usually activation and onboarding rather than anything further down.

If the curve flattens, acquisition compounds. Every dollar adds to a base that persists, and this is the moment to scale channels hard — because the thing that makes spend worthwhile is now demonstrably true.

Most companies get this order backwards, and it is the most expensive sequencing mistake available to an early-stage team.

Questions