A data-driven framework for investigating customer retention

Abstract

This study presents a data-driven simulation framework in order to understand customer behaviour and therefore improve customer retention. The overarching system design methodology used for this study is aligned with the design science paradigm. The Social Media Domain Analysis (SoMeDoA) approach is adopted and evaluated to build a model on the determinants of customer satisfaction in the mobile services industry. Furthermore, the most popular machine learning algorithms for analysing customer churn are applied to analyse customer retention based on the derived determinants. Finally, a data-driven approach for agent-based modelling is proposed to investigate the social effect of customer retention. The key contribution of this study is the customer agent decision trees (CADET) approach and a data-driven approach for Agent-Based Modelling (ABM). The CADET approach is applied to a dataset provided by a UK mobile services company. One of the major findings of using the CADET approach to investigate customer retention is that social influence, specifically word of mouth has an impact on customer retention. The second contribution of this study is the method used to uncover customer satisfaction determinants. The SoMeDoA framework was applied to uncover determinants of customer satisfaction in the mobile services industry. Customer service, coverage quality and price are found to be key determinants of customer satisfaction in the mobile services industry. The third contribution of this study is the approach used to build customer churn prediction models. The most popular machine learning techniques are used to build customer churn prediction models based on identified customer satisfaction determinants. Overall, for the identified determinants, decision trees have the highest accuracy scores for building customer churn prediction models.