Will they stay or will they go? Predicting customer churn in the energy sector

The liberalisation of the energy market has driven changes in the way firms approach marketing, both for the acquisition of new consumers and for retaining existing ones. To retain consumers, practitioners aim to predict which consumers intend to churn (ie leave), and to understand the reasons behind this intention. To address this need, this study uses data-mining techniques to develop a churn prediction model. The study aims to identify the information that is predictive of churn and, consequently, to shed light on the psychological reasons behind churn. The authors built eight predictive models using decision trees, random forest and logistic regression on a dataset composed of 81,813 consumers of an energy provider, each with one residential electricity contract. The logistic regression was found to outperform the other methods. The discussion focuses on the relevant predictors of churn by addressing a posteriori psychological explanations of consumers’ churn behaviour. The study provides new insights on the reasons why customers churn and, by addressing theoretical psychological explanations, provides a data-mining model with robustness to contextual changes.