🔗  Qini Curve

Qini Curve#

An uplift-evaluation curve plotting incremental gain against the fraction targeted.

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 Despite careful review, it may still contain mistakes, omissions, or out-of-date information.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#

The Qini curve is the visual evaluation tool for uplift models (causal or incremental-response models). It shows the incremental benefit gained by targeting customers ranked by predicted uplift score — the uplift-modelling analogue of an ROC curve, measuring how well the model finds persuadables. The higher and more “bowed upward” it is, the better.

How it’s built#

Rank customers by uplift score from highest to lowest, incrementally target the top x*%, and for each portion compute the **incremental response* — treatment responses minus control responses. The cumulative incremental response is what gets plotted.

Axes and baselines#

The X-axis is the proportion of the population targeted (0% to 100%); the Y-axis is the cumulative incremental response. A random-targeting diagonal marks the uplift you would expect from targeting at random, and the model curve ideally sits well above it.

Reading it, with an example#

Targeting the top 20% might show a treatment purchase rate of 18% against a control of 12% — a +6% incremental lift on those customers — and so on at 40%, 60%. A steep early rise means the model concentrates persuadables at the top; a flat, near-diagonal line means it is no better than random. The area between the model curve and the diagonal is the Qini coefficient.


Theme: Causal Inference & Uplift  ·  All terminology



See also

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

Tags: purpose: reference topic: terminology level: advanced