Question 1

What is churn prediction software?

Accepted Answer

Churn prediction software watches how customers actually use a product and how they pay for it, then estimates which accounts are likely to cancel while there is still time to intervene. It differs from cancel-flow tools, which only act once someone has already decided to leave.

Question 2

What is a customer health score?

Accepted Answer

A customer health score is a single number, usually 0 to 100, summarising how healthy an account looks right now. Retain builds it from signals like key feature usage, usage trend, recency and payment status, each weighed by the importance and expected frequency you set for your own events. The number matters less than the direction it moves.

Question 3

How does Retain predict churn?

Accepted Answer

In four steps. First you tell Retain which events matter in your product, how important each one is on a 1 to 5 scale, and how often you expect it. Then the AI reads more than 10 behavior and billing signals per account, including key feature usage, onboarding completion, usage trend, days inactive, trial activation and failed payments. Those collapse into one health score from 0 to 100, recalculated daily and mapped to a risk level. When an account changes risk, an alert fires carrying the factors that caused it.

Question 4

Is this just a "no login in 30 days" rule?

Accepted Answer

A static rule uses one threshold for every customer, so it fires late on a heavy daily user and constantly on a light monthly one. It has no idea what normal looks like for either. Retain reads each account against the frequency you said you expect for your own events, so a daily team going quiet for a week registers as a bigger problem than a monthly reporting tool going quiet for three.

Question 5

How far ahead does Retain warn me?

Accepted Answer

It depends on the signals, not on a fixed window. Scores are recalculated daily, so a drop in usage shows up the day after it happens rather than at renewal. In practice that means weeks of warning on gradual disengagement, and same-day warning on a failed payment.

Question 6

What data does churn prediction need?

Accepted Answer

Two streams. Product usage events, sent through the HTTP API or an SDK, which is what tells Retain whether an account is getting value. And subscription data from your payment provider, synced from Stripe or Polar, which supplies plan, MRR and payment failures.

Question 7

Does it work if I do not have years of churn history?

Accepted Answer

Yes. Retain reads observable behavior against the expectations you set for your own events, so it does not need years of your cancellations before it produces a useful reading. Accounts without enough signal yet are marked as pending instead of being given a confident-looking number.

Question 8

Is this the same as Paddle Retain?

Accepted Answer

No. Paddle Retain, formerly ProfitWell Retain, recovers failed payments through dunning. Retain (retain.so) detects failed payments too, but its main job is predicting voluntary churn from product behavior, weeks before anyone reaches a cancel screen.

Turn customer signals into churn predictions

Retain uses AI to analyze real customer behavior and identify the patterns that precede churn, weeks before anyone reaches a cancel button.

You define healthy.
Retain finds the risk.

You set what normal looks likeWhich events matter, how much, and how often you expect them.

The AI reads 10+ signalsKey feature usage, onboarding, trend, days quiet and billing.

You get a score, and the reasons for it0 to 100, the risk level it maps to, and the signals behind both.

Give your team time to intervene

Churn prediction, answered plainly.

Start predicting churn today

Other Features