Optimize payments made on third-party insurance claims

Client: UK-based personal and automotive insurance provider with businesses across global insurance, investment management, and other financial services

OPPORTUNITY: The client wanted to optimize the payments made on third-party insurance claims in its personal insurance and automobile insurance businesses. The client wanted to improve upon existing solicitor’s negotiation and settlement based process.

SOLUTION:

SGA implemented the following steps to design a customized solution for the client:

  • Regression: To predict if settlement value exceeds COA, the SGA team designed a Logistic Regression model using independent variables such as solicitor score, solicitor location, injured body part, number of people injured, and vehicle type
  • Segmentation: The SGA team filtered the successful claims on the basis of average settlement value, using various target variables as above. The team then designed decision-tree algorithms to determine the actual segmentation.
  • Offer value identification: Using the above tools, the SGA team valued any new claim based on the average settlement value of the successful cases in the relevant segment

VALUE DELIVERED:

  • Delivered 82% accuracy on unseen data
  • Aided growth of 18% in revenue on a sequential basis

Client

UK-based personal and automotive insurance provider with businesses across global insurance, investment management, and other financial services.

OPPORTUNITY

The client wanted to optimize the payments made on third-party insurance claims in its personal insurance and automobile insurance businesses. The client wanted to improve upon existing solicitor’s negotiation and settlement based process.

SOLUTION

SG Analytics implemented the following steps to design a customized solution for the client:
  • Regression: To predict if settlement value exceeds COA, our data scientists designed a Logistic Regression model using independent variables such as solicitor score, solicitor location, injured body part, number of people injured, and vehicle type.
  • Segmentation: Our team filtered the successful claims on the basis of average settlement value, using various target variables as above. The team then designed decision-tree algorithms to determine the actual segmentation.
  • Offer value identification: Using the above tools, we valued any new claim based on the average settlement value of the successful cases in the relevant segment.

VALUE DELIVERED

►
1
Delivered 82% accuracy on unseen data.
►
2
Aided growth of 18% in revenue on a sequential basis.