An insurance company had several consecutive years a composite with rates of churn in their field of self-superior to other competitors in the market and was gradually increasing. The price was not the deciding factor of lower customers as rates overall were quite competitive.

Minimising churn


Insurance company


Development of a predictive model based on algorithms to detect the customer’s pattern behavior in their previous downward months. These data allowed the detection of customers interests and needs in the company to define shares for retention.

Developed solution

A predictive model based on supervised algorithms to detect the pattern of customer behaviour in their previous downward months developed. Historical data were used 18 months. Predictor’s different data sources that allowed evaluate the customer’s relationship with the company from several viewpoints were used: losses, claims, premiums, antique, ….
On customers classified with high probability of default a value clustering allowed to know the type of each of the patterns of behaviour that led the client to unsubscribe performed. From this clustering which retain customer interest was detected by value and how they should be holding the shares.

Results obtained

The churn rate in the field of car not only stabilised, on the contrary it had declined during the following two years. The knowledge acquired shares is allowed for different areas of the company:

  • Creating a new product for young people with conditions that made it competitive, while maintaining positive return
  • Improved customer experience in pursuing claims
  • Customer orientation and not a product. Encouraging customers with policies in several branches
  • Premium reduction effort to high-value customers.

Churn reduction


Claims optimization process


Premium reduction effort