Reading subscription churn by cohort instead of by month

Blended monthly churn of 7% can mean a healthy business or a leaking one. Plot survival by signup cohort and the shape appears: a cliff at shipment two, then a slower slide around month five.

Shipment-two churn is nearly always a product-fit or expectation problem. Month-five churn is boredom, and the fix is assortment rotation, not discounting.

Measure the intervention against the cohort it was aimed at. A win-back offer sent to everyone flatters the average while doing nothing for the cliff.