Modelling Customer Lifetime Value in a Continuous, Non-Contractual Time Setting

The need of online retailers to maintain a competitive advantage in the today\'s booming online retail industry has led to an increased focus on customer relationship management (CRM). The aim of CRM is to increase a company\'s profits by creating long-term relationships with their profitable customers. However, before this can be accomplished, these profitable customers first need to be identified. The profitability of a customer is often expressed in terms of customer lifetime value (CLV), which is the net present value of all future purchases by a customer. The goal of this research is to compare the predictive power of several different classes of prediction models with respect to predicting CLV. These classes include probability models that are specifically designed to model customer purchase behaviour, duration models that model the general time until a customer\'s next purchase, and machine learning techniques. This research shows that, for Winkelstraat.nl\'s database of customer activity, probability models are most suitable for predicting CLV.