Subsidy Incrementality (PSM)

Project Overview

Subsidies are widely used as growth levers in local services, but traditional metrics such as redemption rate and apparent ROI often overestimate their true impact. In this project, I built a causal inference framework using Propensity Score Matching (PSM) to estimate how much subsidized GMV was truly incremental, identify high-response user segments, and translate the findings into a budget reallocation strategy.

Methods

• Applied Propensity Score Matching to construct comparable treatment and control groups from users receiving the same subsidy batch.
• Validated matching quality using standardized mean differences across historical GMV, order count, same-BU spending, and active days.
• Estimated Average Treatment Effects across 3-day, 7-day, and 14-day windows to evaluate short-term lift and persistence.
• Decomposed apparent subsidized GMV into true incremental contribution and estimated non-incremental coverage.
• Compared incremental ROI across user activity tiers, city levels, consumption segments, and business units.

What I Did

• Designed the end-to-end analytical framework from problem definition, metric design, causal evaluation, segmentation, to strategy recommendation.
• Defined decision-oriented metrics including ATT, Lift, incremental ROI, true incremental share, and estimated waste rate.
• Built a PSM-based counterfactual logic to avoid over-crediting subsidies for users who would have purchased anyway.
• Identified high-potential subsidy targets such as mid-activity users and lower-tier city users based on heterogeneous ROI.
• Translated analytical findings into practical actions: user targeting, coupon structure optimization, budget reallocation, and monitoring dashboard design.