Back to Projects
🎯 Targeting & Growth

Up-Sell Ad Campaign for Intuit QuickBooks

UCSD · Customer Analytics · Dec 2025

Analyzed clickstream behavior of 1M QuickBooks users to build an uplift model that identified which users would respond to an up-sell campaign vs. those who would convert organically. The model informed a precision-targeted second wave that generated $380K in incremental profit.

$380K

Incremental profit

1M

Users analyzed

4

User segments identified

Approach

Used causal inference techniques to separate the treatment effect of the ad campaign from organic conversion. Built uplift trees to segment users into persuadables, sure things, lost causes, and sleeping dogs — then only targeted the persuadables in the second wave.

Impact

The second-wave campaign, informed by the uplift model, achieved significantly higher ROI than the blanket first wave by avoiding wasted spend on users who would have converted anyway and users who were unreachable regardless of targeting.

Tools & Technologies

PythonScikit-learnCausalMLPandasClickstream Data

Topics

Uplift ModelingCausal InferencePythonCampaign Optimization
← All Projects