Seeking high-quality third-party data

April 9, 2024 5:00 PM

Deep Inside The Iron Mountain Data Storage Facility — Employees walk along a roadway inside the Iron Mountain Inc. data storage facility in Boyers, Pennsylvania, U.S., on Feb. 13, 2018.

At InsurtechNY in March, Bryan Adams, head of catastrophe analytics at Arch Insurance, spoke on a panel discussing how to improve combine ratios using third-party data. Combine ratios add together loss and expense ratios, and are a measure of an insurance company's profitability. Their importance to insurers means that the data used to calculate these ratios has to be correct and accurate. That data even drills down to imagery used to authenticate details about property being covered, for example. Such imagery is now subject to AI processing to improve accuracy. Digital Insurance spoke with Adams after the conference about how Arch manages third-party data and meets these accuracy and operations challenges.

What is Arch Insurance’s process for dealing with third-party data and choosing sources?

Bryan Adams of Arch Insurance — Bryan Adams, head of catastrophe analytics, Arch Insurance.

One of the difficulties in obtaining third-party data is the amount of it that actually exists out in the ecosystem. It ends up being a little bit more about trying to find data that's fit for purpose, and making sure that we aren't just consuming all types of data from all vendors, that it's really about understanding the value that it brings. And then going out and seeking different types of vendors, maybe vendors in the same space and understanding the strengths and weaknesses of each as we look to bring them in. But it's really in many ways about selecting the right data for the right purpose.

What criteria do you use to select the right data?

It's more specific, as we talk about fit for purpose. For example, in light of inflation and other changes in the economy, for valuation, it may just be looking for data sets that take that into account on a more frequent basis, or bringing in that information so that it's more relevant and more updated. It's keeping track with the time center in place. It really depends on each of the individual datasets, unfortunately, to make sure. But overall, probably timeliness. That it's not stagnant and it's continually updated.

What are the challenges for the insurance industry in handling third-party data?

It's navigating all the different sources that are out there. As we saw at InsurtechNY, there's a lot of vendors selling a lot of data. With insurance, it's hard to always tease out the return on investment on that data. A lot of it is hypothetical in nature. It's trying to understand the problem that you're trying to solve, and selecting an approach or data that helps you solve that problem, and not the other way around. Not letting vendors tell you the problem that they're trying to solve is actually the problem or optimizations that you want to solve for, and finding the data that fits that.

What are some methods or approaches that carriers can use to solve those issues?

Data and analytics, understanding the book, understanding claims, understanding the drivers of claims, understanding any type of perils that a company may write in and having a grasp on what is driving the need for third-party data. It could be on the loss reduction side, it could be on the expense reduction side. But what is the goal of what you're trying to do? Make sure that it's clearly defining what you're trying to define, which will then help in ROI calculations and being able to select data that makes sense.

Is there room for new entrants in this space?

For some of these products, the barrier to entry is quite low. Some of them, maybe not. But there's absolutely opportunity for new startups to get involved in the space and be able to get some traction relatively quickly.

What is Arch Insurance watching out for or expecting to see?

There's two trends I see. One is around AI, more specifically, bringing together multiple different data sources to mine that information in a more generic fashion. We're just scratching the surface on it, as we continue to go forward. The second piece is remote sensing in general, whether it's through higher resolution satellites, higher resolution camera systems on fixed wing aircraft. Then using that information, and drones, to understand the landscape or particular structure at a much higher resolution than before. Those are the two areas that I'm watching out for, to see them come to fruition over the next few years.

Is the quality of images from remote sensing changing?

I've seen this space go through tremendous improvements in frequency and resolution over the past few years. That's really a result of better camera systems. Now after an event happens, within 24 to 48 hours, we have high resolution imagery that's shot one, two, three days in a row after an event. The Maui fires are a great example of having satellites capture that information. Within 48 hours, we knew the full footprint, all the buildings that were damaged. That has gone a long way to help the insurance industry just understand and quantify risk, but also assess each individual policyholder.