🔗 Cumulative Uplift

Cumulative Uplift#

The running incremental effect as more of the targeted population is treated.

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 Despite careful review, it may still contain mistakes, omissions, or out-of-date information.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

Cumulative uplift is the running total of the incremental effect — extra responses, conversions or purchases — gained by targeting customers ranked by uplift score. It says how much total benefit has accrued up to a given proportion of the population, and it is the quantity on the Y-axis of the Qini curve.

The formula#

For the top \(k\%\) of customers ranked by uplift score,

\[\text{Cumulative Uplift}(k) = \sum_{i=1}^{kN} \left( y_i^{\text{treatment}} - y_i^{\text{control}} \right),\]

summing the treated-minus-untreated outcome over those customers, where \(N\) is the total count.

A worked example#

Rank 1,000 customers (split treatment/control) by uplift score and walk down the deciles. The top 10% gives 18 treatment versus 12 control conversions — +6, so cumulative uplift is 6. The top 20% adds 32 versus 24 — +8, bringing it to 14. The top 30% adds +10, reaching 24. At 100% the final cumulative uplift equals the total incremental benefit.

Why it matters, and its cousins#

The curve shows where uplift concentrates (the top 20% may hold most of the benefit), fixes the optimal targeting cutoff (stop where it flattens), and evaluates models — steep early is good, flat like the random baseline is bad. Its family: incremental gain is the per-group step, cumulative uplift (the cumulative incremental gain) the running sum, total incremental benefit the final value, and the Qini curve their plot.

Theme: Causal Inference & Uplift · All terminology

Hint

Mind map — connected ideas

Cumulative Incremental Gain (CIG) · Incremental Gain · Total Incremental Benefit (TIB) · Qini Curve · Uplift Score · Uplift Models

Hint

More in Causal Inference & Uplift

AUUC (Area Under the Uplift Curve) · Causal Effect · Causal Impact · Causal Inference · Causal ML (Causal Machine Learning) · Causal Trees · Cumulative Incremental Gain (CIG) · Incremental Conversions · Incremental Gain · Incremental Recovery Rate (IRR) · Incremental Revenue · Incremental Sales · Qini Coefficient · Qini Curve