Build Predictive Model for Buyer Engagement Uplift
Overview
This question evaluates proficiency in causal inference and uplift modeling, experimental design and sample-size/power analysis, feature engineering and leakage control, model selection for treatment-effect estimation, and translating predicted incremental impact into a launch decision.
Meta
Jul 12, 2025, 6:59 PM
Data Scientist
Technical Screen
Machine Learning
5
0
Predicting Engagement Uplift for a New "Show similar products" Button
Scenario
A new UI control (a "Show similar products" button) may change buyer engagement. You need to predict the causal impact (uplift) of showing this button and use the result to recommend whether to launch globally, target to segments, or not launch.
Task
Describe how you would build a model to predict engagement uplift and support a launch decision, covering data/assumptions, features, algorithms, sample-size/power, evaluation, and decision threshold.
Requirements
- State assumptions and the data/experiment setup you need to estimate causal impact (not just correlation).
- Propose feature sets (demographic, behavioral, contextual), with leakage controls.
- Choose algorithms appropriate to data size and outcome type, including uplift/causal options.
- Explain sample-size and power considerations for ATE and heterogeneous treatment effect (uplift) modeling.
- Define evaluation methods and metrics for uplift models and how you would cross-validate.
- Specify how to set a decision threshold and translate predicted uplift into a launch or targeted rollout recommendation.
Hints
- Discuss demographic and behavioral features.
- Pick an algorithm based on data size and sparsity.
- Cross-validate and use uplift-aware metrics.
- Translate predicted uplift into a launch recommendation (global vs targeted) with business constraints.
Solution
Login required
Show