Understanding Users' Coupon Usage Behaviors in E-Commerce Environments

Issuing discount shopping coupons is a popular way to promote sales in E-commerce environments. However, it is unclear that whether a customer will use the coupon or not. In order to improve the coupon usage ratio, it is important to predict the probability in which a user will use a coupon. In this paper, we treat the coupon usage probability prediction problem as a binary classification problem, and we use machine learning methods to analyze users' coupon usage behaviors. We conduct extensive data analysis on users' coupon usage behavior data, provided by Alibaba in Tianchi platform. Our method gains a good prediction result (with the AUC of around 0.78). Besides that, we find some important factors/features that can significantly impact users' coupon usage behaviors. We also find the ”old user effect.” Moreover, we find some overfitting features that have little effects and remove them. Those findings can guide merchants to issue proper coupons to proper customers.