A Price Driven Hazard Approach to User Retention

TL;DR

This paper introduces a hazard-based model utilizing Cox's proportional hazard approach to predict user retention by analyzing interaction data, considering price sensitivity, user-product alignment, and external influences, demonstrating superior performance on real data.

Contribution

It presents a novel hazard model for user retention that integrates multiple factors affecting purchase intent, improving prediction accuracy over existing methods.

Findings

The model outperforms state-of-the-art algorithms in real-world tests.

Incorporating covariates enhances the model's predictive power.

The hazard approach captures dynamic user retention behaviors.

Abstract

Customer loyalty is crucial for internet services since retaining users of a service to ensure the staying time of the service is of significance for increasing revenue. It demands the retention of customers to be high enough to meet the needs for yielding profit for the internet servers. Besides, the growing of rich purchasing interaction feedback helps in uncovering the inner mechanism of purchasing intent of the customers. In this work, we exploit the rich interaction data of user to build a customers retention evaluation model focusing on the return time of a user to a product. Three aspects, namely the consilience between user and product, the sensitivity of the user to price and the external influence the user might receive, are promoted to effect the purchase intents, which are jointly modeled by a probability model based on Cox's proportional hazard approach. The hazard based model provides benefits in the dynamics in user retention and it can conveniently incorporate covariates in the model. Extensive experiments on real world purchasing data have demonstrated the superiority of the proposed model over state-of-the-art algorithms.