Case Study
Model-Assisted Claims Assessment & Decisioning
Challenge
Identifying roof damage can be very time consuming and dangerous for property insurance adjusters. In cases where direct access to a damaged roof is not possible or is extremely unsafe, aerial photos are used to assess the damage. A system to assess roof damage more efficiently was needed In order to provide policyholders with rapid and accurate claim payouts.
Industry
Financial Services / Insurance
Data Type
Image (jpg, png)
Project Duration
3 Months
Ongoing?
No
Solution
To assemble this system, thousands of roof images had to be reviewed and appended with specific damage labels to provide training data for a model to consistently identify roof damage. Insurance adjusters were not available to provide a significant amount of time to perform this data annotation. The solution was to use model assisted labeling, which enhances the initial label output from an automated process with a review by a subject matter expert. A human-in-the-loop process leverages the efficiency of an initial machine step to reduce the time demands on the subject matter experts yet ensures that all output benefits from their experience and judgment. Corrections and edits made by the subject matter experts feed back into the automated process to continually improve the accuracy of the initial labeling
Outcome
The accurately labeled rooftop images powered additional automated steps in claims processing, resulting in more accurate and faster payments to policyholders, while minimizing time demands on subject matter experts
Download Case Study
