Predicting Insurance Churn to Reduce Clawback

Many insurance companies still depend on agents for marketing and selling their products. The compensation for these agents is based mainly on commissions and hence such agents try to sell as many policies as possible. In addition to increasing the quantity of policies sold, agents may try to recommend policies with high commissions even if other policies are more suitable for the client. Furthermore, agents may at times ill-advise customers to change their policies in order to gain a new commission even if such a change does not benefit the customer or the insurance company. This last type of transaction is called internal insurance churn or policy churn since the customer remains with the insurance company but they make a policy change. We use Machine Learning techniques to predict policy churn prior to payment of the commission to the agent. This avoids the process of having the agent repay the commission (a process termed clawback) in such cases. We illustrate how the proposed approach can provide significant savings to the insurance company.