Search
  • Robert Kozikowski, CPO

Recent Devastating Tornadoes: How TensorFlight is Helping

Updated: Dec 31, 2021

Introduction

A devastating series of extreme weather events recently struck the southeastern U.S., including two tornadoes that affected Kentucky, Tennessee, and several neighboring states. These storms occurred between December 10th and 13th, 2021. Our hearts go out to the victims and their families in these trying times. We are developing technologies that can help property owners and their insurers better manage risk in two ways. First, by assessing which structures are most likely to be damaged so that action can be taken to improve the safety of these structures. And second, by facilitating rapid claim adjustments with satellite, aerial, and other imagery. In this article, we highlight the ways in which these innovations will better equip the insurance industry to respond to, and reduce losses from, weather events such as these.


Claims

Rapid damage estimation after a catastrophic weather event is important for all parties concerned, especially property insurers. Claims adjusters need to be able to focus on assisting those that were affected, such as helping them to find temporary accommodations. Insurers also need to be able to set reserves, determine whether external assistance is needed with the adjustment of the claim, and plan for reconstruction costs.


This is where TensorFlight can help. As soon as we know which property suffered potential damage, we can estimate the damage without having to send “boots on the ground.” We scan all available imagery sources, including satellites, aerial imagery, governmental data, and more. Then, we automatically establish whether the image is of sufficient quality, factoring in such variables as cloud cover, and estimate the damage to the structure. In cases where the property is not visible through the clouds, we offer an industry-leading solution, using SAR satellites that can pierce cloud cover and detect buildings from SAR data - https://spacenet.ai/earthvision2020/.


We are not meteorologists, so we can’t forecast weather events. Thankfully, there are excellent solutions in the industry. Our clients recommend the Corelogic weather team to provide forecasts in the affected areas for any given weather event. For more information, contact curtmcdonald@corelogic.com or tmaloney@corelogic.com .


Once you have obtained a shapefile with potentially affected areas from a firm like Corelogic, you then notify us which of your properties may have been affected. TensorFlight will scan all imagery of each property that is available from public and subscription sources. As soon as we obtain a reliable image, we update you with the damage estimation. For example, below is imagery from the recent tornado, obtained from Maxar WorldView satellites on our platform:


Underwriting

Property claim monitoring is an important service we offer, but our primary mission is to provide you with better data at the point of underwriting and risk modeling. Our data is based upon cutting-edge artificial intelligence (AI) and computer-assisted visual research, analyzing multiple images of each property and additional external data sources. Having access to this data will enable you to better estimate the probability and magnitude of a loss. Our datasets include accurate geocoding, number of stories, square footage, replacement cost, construction type, occupancy type, roofing, facade attributes, and more. If your loss development following these tornadoes is more adverse than you had expected, it may be time to revisit the property data you received from your broker several years ago.


We analyzed many of the properties that were affected by these tornadoes. We are sharing with you a TensorFlight property dataset as seen before the tornado. The dataset we are sharing is generated by our fully automated algorithm, and this data is accessible to users via our API (https://app.tensorflight.com/assets/docs/api_documentation.html). We have more addresses in our database, but we only shared a subset as permitted by our licenses. Please check if the property you are currently adjusting is in the database and compare our data with your records. This dataset is based upon a fully automated algorithm derived from AI analysis; data we provide commercially to our clients is selectively reviewed by our team of property analysts, using confidence scores provided by the algorithm and the property value.


Here, you can find a spreadsheet with all buildings in the path of the recent tornado: https://docs.google.com/spreadsheets/d/1Sz14LnelN7p2N-0qnhvNLm-y0VHpDCjhQlI6vR8woF0/edit


Conclusion

TensorFlight is working hard to help the insurance industry respond to events like this – with greater insights for predictive modeling before the event and actionable intelligence after the event. Our goal is to bring greater precision to the underwriting process and more efficiency to the adjustment of property claims. Please get in touch with us at contact@tensorflight.com.


660 views0 comments

Recent Posts

See All