Reactivation model

The probability that a churned customer comes back, used to target win-back efforts

What it predicts

For customers who have churned, the probability that they start a new subscription or agreement again. The probability is shown as a reactivation score from 1 to 10 and a reactivation level (High, Medium, Low), the mirror image of the health scoring model but for lapsed customers.

Data it needs

The subscription history of churned customers (how long they stayed, how they cancelled, whether they reactivated before), the time since they became inactive, their profile, and any engagement they still show such as email opens or website visits.

How it works

The model is trained on past churners and whether they came back within a set window, using the same model families as the health scoring model. Time since last active is a dominant feature: the probability decays with every month of inactivity, but the decay differs strongly by acquisition channel, product and cancellation reason, which is what the model captures.

How often it runs

Daily for customers who churned in the recent past, typically the last 12 to 24 months.

How to read the output

A high reactivation level means a win-back action is worth its cost; a low level means the customer is best left to low-cost automated touches. The Reactivation page shows how many churned customers were reactivated, the reactivation rate, and how many are still inactive.

Where it surfaces

The Reactivation page, the customer page of churned customers, the High Reactivation Potentials audience, and Co-pilot win-back use cases. The reactivation model is an add-on.