Customer Segmentation for improving Marketing Campaigns in the Banking Industry

Abstract

The internet has had a significant impact on financial institutions by allowing customers to access many bank services virtually thus creating a very competitive environment. Therefore, efficient customer segmentation is a key objective for achieving more profitable market penetration. We propose a hybrid model that predicts a financial institution client’s propensity to transition to an online banking platform. In this research, we utilized a hybrid approach where the first stage is Transaction Cluster Analysis where Recency, Frequency and Monetary (RFM) segmentation and K-Means cluster analysis were performed to detect the most loyal market segments. Analytic Hierarchy Process (AHP) was used to deduce the weightings of each cluster which aided in calculating the Customer Lifetime Value (CLV) of each cluster. Then two clustering algorithms, K-Modes and K-Means, were utilized on the clients’ demographic features. In the final stage, we performed experiments that compared two supervised learning algorithms, Decision Tree and Extreme Gradient Boosted (XGBoost), to predict online transition behaviour. Our results showed that K-Modes clustering algorithm and XGBoost classification model yielded the best test accuracy of 96.1%. Our results illustrate our claims by showing that the bank can attract more customers, maintain its customer base, and keep their important customers satisfied.