Predictive Modelling and Strategies for Customer Churn Reduction: Loyalty Program Case

Effective marketing strategies to retain customers are crucial in today’s competitive business landscape. Churn or customer attrition, is when consumers or users stop using a particular service or product, which presents a substantial challenge. High churn rates not only result in significant revenue loss but also necessitate Intensified marketing expenditures. Addressing this challenge involves the proactive prediction of potential churners, enabling businesses to implement targeted retention actions and mitigation strategies. This thesis dives into the analysis of user characteristics and behavior within a loyalty program, with the objective of constructing a predictive churn model. The ultimate aim is to refine marketing strategies to boost customer retention. To accomplish this objective, a dataset encompassing diverse customer data, including demographics, access and upload histories, and product descriptions, was collected and analyzed. The preparation of the data was a crucial step before the implementation of various classification algorithms. The final model achieved 90% accuracy with a balanced F1 score. Furthermore, users were categorized into four distinct risk levels through the utilization of a propensity score, ensuring a more targeted approach to retention strategies. Moreover, to get a maximum understanding of the differences between and within the levels, another segmentation was implemented using a surrogate model. This thesis does not only predict customer churn but also places a spotlight on the customization of communication as a central strategy. The ability to categorize users based on their risk levels empower businesses to tailor their communication approaches, ensuring a more personalized interaction with their clientele.