Cluster Analysis for Customer Segmentation with Open Banking Data

Segmenting customers into different groups using their characteristics and behaviors has always been an important topic. Customer segmentation can lead to better customer understanding and targeting, which in turn leads to more effective product tailoring and marketing strategies. Data mining methods are powerful techniques that can be used in customer segmentation to find customers with similar characteristics. Past research that evaluated different data mining techniques has often had drawbacks, such as using too time-consuming methods or conducting studies with smaller data sets. Density-based clustering algorithms for customer segmentation, such as the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) has only been examined by a few research papers. This study, which summarized the main findings of the unpublished dissertation of Bartels [2021], aimed to classify the segmentation of customers using a Recency, Frequency and Monetary Value (RFM) Model and the clustering techniques, K-Means and DBSCAN, to find groups of similarities and differences and to discover potential valuable and vulnerable customers. The data used was from Open Banking data sets, including anonymized transactions from various bank customers in the UK for three months in 2017 and mainly focused on different types of expenses. K-Means found three clusters each month that represent the most, medium, and least valuable customers. While the most valuable customers have the highest average values per attribute, the least valuable customers have the lowest average values. The found clusters were analyzed and evaluated to find potential vulnerable and valuable groups, which can help with future product tailoring and marketing, especially for unforeseen emergency circumstances such as a pandemic. K-Means outperformed DBSCAN, as the latter showed negative silhouette coefficients.