🔗  Uplift Score

Uplift Score#

A model’s estimate of how much an action changes an individual’s outcome.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

An uplift score is the number an uplift model assigns to an individual, estimating the incremental impact — positive or negative — of a treatment (a campaign, discount or medical intervention) on that person’s likelihood of acting. It answers: how much more (or less) likely is this person to act if we intervene than if we do not?

The formula#

At the individual level,

\[\text{Uplift Score}_i = P(\text{Outcome} \mid \text{Treatment}, i) - P(\text{Outcome} \mid \text{Control}, i).\]

A positive score means the treatment raises the chance of the desired outcome, a negative score means it backfires, and a score near zero means the treatment barely matters.

A worked example#

For a subscription campaign: Customer A has a treated probability of 0.40 versus 0.25 untreated — an uplift of +0.15, a strong target. Customer B sits at 0.70 versus 0.68 — just +0.02, barely worth the spend. Customer C is 0.10 versus 0.20 — an uplift of -0.10, meaning the campaign actively hurts (a “Do-Not-Disturb” case).

How it’s used#

Uplift scores let you rank and target the highest-scoring customers — the persuadables — while avoiding sure things and lost causes, and excluding negative-uplift customers who might churn or unsubscribe if contacted. Aggregated across the population, the scores build the uplift and Qini curves whose quality the Qini coefficient summarises.


Theme: Causal Inference & Uplift  ·  All terminology



See also

Source article Adapted (context, re-expressed) in our own words from: Uplift Score (insightful-data-lab.com).

Tags: purpose: reference topic: terminology level: advanced