Churn Model

Details

Supervisors

Martina Vandebroek

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

AXA Bank Belgium (ABB) has experienced an increase in churn caused by changing customers’ preferences and regulations meant to render switching banks easier. Simul- taneously, attracting new customers appears to be cumbersome and expensive, while retaining its customer base offers promising outcomes. This led ABB to dedicate more importance to anticipating the loss of a customer, and understanding why customers might decide to churn. Although research on churn is plentiful, the actual challenge lies in devising a bridge between the existing theoretical concepts, and operational data and processes. This project therefore aims at conducting a comprehensive churn analysis for ABB, with as starting point the creation of a theoretical background on churn through a literature review, and with as main outcome, a churn model. Using the data provided by ABB, this project creates a completely new data set for predicting churn, and performs the usual exploration, cleaning and pre-processing steps for modelling churn. The results show that the determinants of churn identified in practice do not completely coincide with those in the literature, and behavioural characteristics seem to take the lead over demographic ones. Additionally, while tree-based ensemble methods achieve the best performance, they do not significantly outperform logistic regression which also offers an enhanced interpretability. Eventually, restricting the study to active forms of churn enables to formulate a baseline churn framework, while including more passive cases of churn broadens the scope and offers promising outcomes for ABB. The achieved results trigger an interest to further explore other churn definitions.