Freemium: Uplift Modeling for Marketing

Hello!

Welcome to today’s edition of Business Analytics Review!

Here’s an uncomfortable number: at most retailers running discount campaigns, somewhere between 40% and 60% of “converted” customers would have bought the product with or without the coupon. The response model that flagged them as high-value targets was right about their behavior and completely wrong about the campaign’s effect on it. Marketing teams have spent two decades optimizing models that answer “who will buy,” when the only question that pays the bills is “who will buy because we asked.”

This isn’t a rounding error. It’s the difference between a campaign that prints money and one that quietly subsidizes purchases that were already locked in. And it explains why so many teams with excellent classifiers, strong AUC scores, and clean dashboards still can’t explain why quarterly revenue didn’t move when the campaign “worked” by every metric they tracked.