Property Insurance Claims estimation

The insurance claim estimation process can be automated by machine learning algorithm. Computer vision has the capacity to look at high resolution images to reconstruct the property and measure roof parameters. Through supervised training, it can also look for damaged areas in all of the images.

Problem

Insurance claims are very long processes, especially when it comes to large properties such as residential houses. The standard procedure when someone files for a claim includes: inspection, estimation, and repair.

Each step can take as long as a few weeks due to the inefficiency in organization and manual works. This is particularly true during hail or thunderstorm season when large communities of houses are damaged at the same time. Many insurance companies are overwhelmed by the claims and unable to respond timely. Their customers get increasingly frustrated and affect customer satisfaction in a negative way.

Inspection requires a professional roof inspector to go to the property location and climb onto the roof of the house. He then needs to check all areas of the roof and look for damages. He needs to use a camera to take pictures of the damaged spots.

This process usually takes 1-3 hours depending on the size, height, and slope of the roof; it not only is manually intensive, it also poses danger to the professional where falling off the roof could realistically happen and cause severe bodily harm.

The estimation process requires a roofing professional to look at the pictures taken during inspection, and based on the damaged areas, roof type, shingle type, etc. to estimate how much it will cost to repair the roof. Again, it is a manually intensive process that could take hours per case to complete.

Solution

The inspection process can be replaced by a drone. Drones nowadays have the capability to fly over a house and collect images through high resolution cameras. They are fully controllable by certified drone pilots and poses minimal danger to the operators.

Taking 30-40 images around the roof usually takes less than half an hour and it reduces the manual work required by the process.

The estimation process can be automated by machine learning algorithm. Computer vision has the capacity to look at high resolution images to reconstruct the property and measure roof parameters. Through supervised training, it can also look for damaged areas in all of the images.

This eliminates the necessity for all the manual works in the backend and dramatically increase the automation and efficiency of the work flow.