The diversity of products and the lack of transparency make it difficult for consumers to come to a decision. The biggest challenge in advi‑ sing the customer and recommending a complementary insurance is the incomparability of the products. Providers design their policies and bundle the benefits in such a way that comparisons are virtually impossible. Comparis was on the lookout for new strategies to help customers select the right insurance. They should not have to rely solely on product information, but also receive recommendations for suitable products based on their user profiles.
Zuhlke evaluated the potential of data-driven approaches for optimizing the customer journey. Latent class analysis of historical user data enabled the team to automatically identify typical patterns in the process of choosing an insurance company. This resulted in well-defined types of insurance policyholders. Several classes were identified and linked to the user profiles through a self-learning Random Forest algorithm. With the models derived in this way, Comparis gained new insights into the users’ behavior and was subsequently able to pre‑ sent the customer with information like: «Other interested parties in a similar situation selected product A, B or C with an x, y or z percent probability.»