Comment by boscillator

What is the point of predictive analytics for them? Surely they don't care about what one specific customer will do, only in aggregate? Since ML models can only learn from past aggregates, wouldn't traditional statistics on past transactions make more sense?