Propensity model

A propensity model estimates the probability that a given customer will do something within a defined window — buy again, lapse, upgrade, open. Lifecycle teams use the scores to target, sequence and suppress, so a poorly built model quietly misdirects the whole programme. Two things matter more than the choice of algorithm: a clean definition of the outcome and its time window, and scoring only people who were genuinely eligible. Train on everyone and the model mostly learns who already buys a lot.