From data acquisition to validation: a complete workflow for predicting individual customer lifetime value

Customer lifetime value is a core measure that allows companies to predict the potential net profit from future relationships with their customers. It is a metric that is computed by recording customer behavior over a long term and helps to build customized business strategies. However, existing research focuses either on a conceptual model of customer $${\text {CLV}}_{\text {s}}$$ CLV s or assumes that all variables required for the computation of CLV are readily available. In this research, we employ a real customer dataset of insurance policies to construct a holistic framework that covers all aspects of CLV computation. In addition, we develop an extensive validation process, aiming to verify our results and obtain an understanding as to which CLV models perform best in the insurance context. In this research, we deliver a framework which comprises all aspects of CLV estimation using a real insurance policy dataset provided by a large business partner. The framework addresses the creation of a unified customer record, classification of customers into ranked groups, interpolation of missing parameters, through to the calculation and validation of individual CLV values. Our method also includes a robust validation with both subjective and objective evaluations of our findings.