Prediction of Customer Churn in Insurance Industry (Case Study)

With the saturation of markets, the organizations have found that retaining customers- especially valuable customers- should be at the center of their management strategies because the cost of attracting new customers is more than the cost of retaining existing ones. The insurance industry is no exception and because of low switching costs is faced with customer churn. This study examines the impact of length of relationship, time of purchase, frequency of purchase, monetary value, profitability and group of purchased products on the valuation of customers. To do this, a survey is used to elicit expert opinions on the effective variables in the valuation of customers. Results indicate that the variables of purchase frequency, length of relationship and the number of groups of purchased products are most important in valuation of customers from the perspective of experts. Then, we develop a model of customer churn to determine the factors that affect high value customers’ churn. Using different methods (neural network, decision tree, logistic regression and support vector machine), predictive models are built and their accuracy is evaluated. The results show that C5.0 decision tree model has a higher accuracy and precision than other models in churn prediction.