[5Yin-A] Poster 8

[5Yin-A-01] Evaluating CRM Campaign Effects with Distributional Treatment Effects

〇Masahiro Tanaka¹, Asuto Hirano¹, Shota Yasui¹ (1. CyberAgent)

Keywords:

CRM,Causal Inference,Causal Machine Learning,Customer Relationship Management,Distributional Treatment Effect

In Customer Relationship Management (CRM), A/B testing is widely used to evaluate the effectiveness of campaigns, and analyses are typically based on the Average Treatment Effect (ATE). However, outcomes in CRM, such as sales amounts, often exhibit mixed discrete and continuous structures and are characterized by zero inflation due to many customers not responding to campaigns. Under such data characteristics, evaluations based solely on average treatment effects may fail to capture the distributional structure of campaign effects and their impact on specific subsets of customers. This study demonstrates that focusing on Distributional Treatment Effects (DTE), which consider the entire outcome distribution, enables a clearer understanding of CRM campaign effects. Through a concrete data analysis example, we show that DTE can visualize effect heterogeneity and asymmetry that are overlooked by average-based evaluations, providing an interpretable and complementary perspective for CRM campaign evaluation.