Churn prediction is the use of customer data and statistical or machine-learning models to estimate which subscribers are likely to cancel in the near future. Its purpose is to warn retention teams early enough to act, since a customer is usually easier to keep than to win back. The output is typically a risk score per account, updated daily or weekly, that feeds customer retention work.
Common signals include falling log-ins or usage, fewer active seats, unresolved support tickets, negative survey scores, approaching renewal dates and payment problems such as expiring cards or repeated declines. Models range from simple rules to logistic regression and decision tree ensembles, often built with behavioral analytics data. Predictions are usually split by cause: involuntary churn from failed payments is handled with pre-dunning and card updates, while voluntary customer churn calls for outreach, training or discounts. Accuracy is limited when history is short or the product changes quickly.