Predict policy level lapses for insurance provider

Client: Leading South East Asia-based insurance provider catering to small and medium businesses

OPPORTUNITY: The client wanted to develop a model to predict lapses at the policy level. The aim of the model was creating cost-effective interventions to target high-value customers with high likelihood of lapse.

SOLUTION:

SGA decided to design a classification model to predict probability of lapse, using the following approach:

• SGA created a data mart that combined demographic data with policy characteristics
• SGA developed a logistic model to predict the probability of a customer lapse on premium payments using predictors such as age, gender, marital status, income, household size, mode of payment, frequency of payment, timeliness of payment, etc.
• The team also designed other models based on Kernel based SVM Regression and Tree based Regression techniques
• Finally, the SGA team calculated the lifetime value index for all the customers. Based on the lifetime value, SGAidentified the customers that should be targeted for business growth.

VALUE DELIVERED:

• Achieved 85% accuracy on unseen data, with an MCC score of 71
• Reduced probability of lapsing from 16% to 7%