Predicting Customer Churn for Insurance Data

Most organisations employ customer relationship management systems to provide a strategic advantage over their competitors. One aspect of this is applying a customer lifetime value to each client which effectively forms a fine-grained ranking of every customer in their database. This is used to focus marketing and sales budgets and, in turn, generate a more optimised and targeted spend. The problem is that it requires a full customer history for every client and this rarely exists. In effect, there is a large gap between the available information in application databases and the types of datasets required to calculate customer lifetime values. This gap prevents any meaningful calculation of customer lifetime values. In this research, we present an approach to generating some of the missing parameters for CLV calculations. This requires a specialised form of data warehouse architecture and a flexible prediction and validation methodology for imputing missing data.