Q&A with John Drzik
The capacity to see changing risk and more fully understand behaviors now exists, yet only a handful of commercial insurers and brokers have started to access, interpret and apply those insights to their business models.
The Council’s Cheryl Matochik sat down with John Drzik, President, Global Risk & Digital at Marsh, to discuss why and the opportunities real-time risk management technologies present the industry.
There are a few reasons. There are new data streams that didn’t exist before. You may have had telematics devices, but they’re now much better and more ubiquitous. So there’s new data coming from a variety of sources such as wearable devices, property sensors, and drones. Equally importantly, AI and machine learning have progressed significantly. They’ve been around for pattern recognition for 15-20 years or more, but are now powerful enough to process the stream of real-time telematics information. I think the combination of different data streams and more live data streams, and the progress of analytics technology, has allowed for the transmission of an active signal of risk in a meaningful way.
I think the industry still has a ways to go processing that signal into a moving risk index. There are a lot of companies now that are starting to do that in certain spaces, where they can take telematics information and give you an expected claims estimate for a driver if they keep driving the same way. Uses like that are what you actually need to make it relevant for insurance. I think we’re only at the beginning of that kind of conversion.
Well, it depends on the line of business. In some, like personal lines, we’re there. To get into commercial property/casualty lines, I think it’ll take a little bit of time to develop. You have to be able to, again, build the convincing case that these analytics are predictive of loss, and that if you use them, you’ll be able to improve your underwriting results.
I think there’s more and more activity in the space, more devices being deployed, more AI models being built, and more insurance carriers that are starting to look at them as underwriting tools. As that evolves, it’s a natural evolution to go to a connected policy where if you change your risk profile, you should be able to get a better insurance price. If you’re going to have a moving indicator of risk influence the pricing – because as an actuary, or underwriter, you’re trying to estimate based on the factors you have – what’s the expected loss from this particular policy? If you can actually get a moving view of the value through the course of the year – and not just once a year, ideally, you would like to change your price in response to the risk change. You have to design a policy to do that.
This has only been done in a limited way in lines like personal auto where people get a credit for good driving, or get some kind of basic change in premium based on their driving results as judged by the telematics. But you could apply that thinking to virtually any line of insurance, if you have what is a concrete and agreed upon metric that is a moving measure of risk.
Statistical models based on demographic information have a high risk of using some variable that’s against the law and is discriminatory. But if you’re judging risk and pricing based on driving behavior, you can evaluate a driver of any race, gender, or age and the telematics doesn’t know that information. The models just know whether a driver is hard braking or accelerating or driving badly. That seems to be a more fair way to judge.
Auto insurance risk is based on behavior, so that’s all that’s going into your pricing. I think using behavioral data is a much, much fairer way. Probably young men are the worst driving risks, right? You’re an 18-year-old boy, you likely won’t get a good price on your auto insurance. But what if you are the 18-year-old boy that actually is a good driver, just as good as the middle-aged guy that you’ll be someday? You can’t convince an auto insurance company to give you the rate of the 55-year old. But if telematics actually show that your driving behavior is exactly the same as really good drivers, then why shouldn’t you get a better rate? Similarly, just because someone is older, doesn’t make them a better driver. So you have to say, you know what, let’s judge that. It may well be that on average, the 18-year-old is a worse driver than the 55-year–old, every statistical actuarial table will tell you that. That doesn’t mean every one of the people in those two pools is the average, even if it’s a fair average.
And that’s where I think that behavioral information has the risk of removing discrimination in the process. If you’re a bad driver, you should get a higher auto insurance price, and if you’re good driver, you should get a better price. That should be based on behavior. Real-time data allows this kind of signaling totally independent of the demographic variables that have been used to price insurance in the past.
Brokers need to move to be more of an advisor to our clients around risk holistically. I’d say historically, we would have advised clients on how they can get the most effective risk transfer, best product, best pricing on their insurance. But let’s stay with the auto insurance example. Let’s say you’re running a small fleet on a commercial auto insurance policy. A broker could help you place that policy and has been doing that for years. But now, you could actually advise on the company itself installing telematics to look at the behavior of its drivers, providing a way to actually reduce the risk in the first instance.
I think that’s really the opportunity for brokers — advising clients on risk mitigation – because who wants auto accidents, right? You do want to be insured against them, but if you could actually do something to limit the number of accidents to begin with, that would be the best outcome. It should lower your claims through time. You take down the nature of the risk exposure. The brokers who become familiar with the arsenal of tools that are now out there to help a client actually monitor and mitigate their risk, and then bring them to the insurance markets with a potentially lower risk, are providing something very valuable for clients. And that cuts across insurance lines. You’d rather not have your building burned down from a fire, you’d rather not have your workers get injured on the work site. If you can just cut that back and use these tools to do that, you’re actually improving safety and risk management before you even get to insurance. And I think that’s really where, as risk professionals, we can help.