Increasing revenue by predicting the percentage likelihood of churn

.

We built a logistic regression model to predict the % likelihood of churn for each individual client based on data held by the insurance group, such as: client tenure, policy count in the renewal chain, product types, premiums, commissions and more. The model ranked clients by highest likelihood to churn so that execs could prioritise their accounts for outreach. Model explainability also indicated potential reasons for the predictions.

We estimated that a 1% reduction in churn for the insurance group would increase revenue by up to 1 million. Based on this, we provided a productionisation plan to bring the POC into live use, and conducted workshops to understand the likely change management activities required. As part of the productionisation plan we suggested how they could continue to enhance the tool with unstructured data.