Churn Prediction

A company wants to know who will leave. The treatment is the decision, and a score on the churn label does not say what to do.

That is the argument of Don't predict churn, Gerben Oostra's "For effective treatment of churn, don't predict churn". In the business to consumer market there are two ways to grow market share, gaining new customers and retaining existing ones, and the second is preventing churn, which makes churn rate, or better retention rate, a key performance indicator; the post asks what a good machine-learning approach to that challenge actually looks like.