Traditional targeting finds people likely to buy. Uplift models find people likely to buy ONLY IF treated. It avoids wasting money on "Sure Things" (would buy anyway) and "Lost Causes" (won't buy regardless).
"Who are the persuadables?"
"Are we wasting coupons on loyalists?"
"Who are we actually annoying with ads (Do Not Disturb)?"
It uses two models (Treatment vs Control) or one unified model to predict the Difference in probability to buy caused by the marketing action.
The Problem
Eroding margin with unnecessary 20% offs.
What It Reveals
Users who need the nudge vs users who don't.
Decision Enabled
Suppress offers for the 'Sure Things'.
The brain behind the Discount Addiction Index. It segments your customer base by incremental response probability.
A Qini Curve showing how much incremental revenue you get by targeting the top X% of uplift-scored users.