Business clients´segmentation based on activity : a banking approach

Abstract

Clustering algorithms are frequently used by companies to segment their customers in order to develop accurate marketing strategies. The K-means is one of the most popular algorithms, despite its drawbacks in terms of seeds’ generation. In this study, several clustering algorithms were tested but in the end the K-means initialized with random seeds was used to segment the data due to its better performance. This B2B segmentation resulted in four clusters based on the activity patterns of each business client, The Loyals, The Minglers, The Challengers and The Believers. Each one of these clusters shows a different type of relationship with the bank, being the bank the first choice for The Loyals and for the Believers but not for the others.