Machine Learning Gives Power to the Underwriting Process for Major Insurance Company
02 November 2018
Our client, one of the largest insurance companies in the world, wanted to double the size of its life insurance business without having to increase its number of underwriters.
DataSparQ created a Machine Learning solution to automate parts of the application assessment in order to increase the number of applications its underwriting team could process.
Most of the life insurance applications that are processed are relatively straightforward and as such, 65% go through the Straight Through Processing (STP) system, which automatically assigns premiums.
The remaining 35% of applications are complex cases that need to be investigated by underwriters. This involves examining additional data from medical records and other sources, as well as information provided by the applicant in free text fields.
Machine Learning Provides a Win Win Solution
Our Machine Learning solution automates the process of assessing the 35% of complex cases in order to identify which of those can be processed via the STP system.
Our solution identified that 9-17% of complex applications were eligible for STP. That means up to 82% of all applications can go through STP, and the number of complex applications that the underwriters have to manually investigate has been reduced, and at its lowest, is just 12%.
Additionally, our client had previously accepted a 1% error rate. An error being an application that should have been declined, but had been underwritten. Our Machine Learning solution reduced the error rate to 0.1%.
Not only has the solution reduced the cost to the business by decreasing the number of errors, but it has enabled the business to increase the number of applications it can process without having to increase its number of underwriters.
Explore another DataSparQ case study for the insurance industry:
We’d love to speak to you more about your challenges and how our data science solutions for the insurance industry can help. Get in touch.