Predictive models for customer attrition

点击翻译按钮获取中文摘要

Research and development in the area of customer attrition or churn/defection is very new and not fully explored yet. More investigation is needed in order to enable businesses to take full advantage of artificial intelligence/machine learning (ML) capabilities to produce a qualitative causal analysis of the potential customer churn. Knowing the causal factors behind the churn can enable businesses to dramatically increase the customer retention efforts, thus improving return on investments. Our goal is to build models capable of predicting the effect of possible customer retention efforts on customer behaviour. Our industry partner, Predikat Inc., focuses on Software as a Service (SaaS) businesses for which the churn rate is among the main indicators of the health of their business. This research project will be determining what techniques to apply to this specific area. We will explore further research opportunity in developing extensions and optimizations of current ML techniques to better solve problems in this domain. In particular, we will design and test a research prototype that will include the theoretical framework and uplift modelling for causal analysis of the customer attrition, as well as effective churn mitigation techniques.