|Ideally, underwriting is based on precise data about risk, gathered over years or even decades, but what happens when the world isn’t ideal? It certainly hasn’t been, as all sorts of anomalies have arisen: COVID, the Russian invasion of Ukraine, supply chain disruptions, inflation and more. To get a better handle on how insurers can underwrite risk in such an uncertain world, ITL talked this month with Henry Kowal, director, outbound product management, insurance solutions, at Arity, an Allstate subsidiary company that tackles underwriting uncertainty with data, data and more data about driving behavior gathered via telematics.|
Auto seems to be a space where there’s even more uncertainty these days than in others. What are gas prices going to be? What's driver behavior going to be like? Will people return to offices and commute like they did before? And so forth? How do you underwrite when things are changing so much?
It's really challenging. A few years ago, I'd never really heard of chip shortages, or supply chain disruptions, at least when it came to auto insurance. Now, the issue is front and center.
But we have lots of data. We just celebrated a milestone this past November, where Arity surpassed 1 trillion miles of driving data. And that driving data is available across the U.S., across states and even down to the ZIP code level. What we see is that total miles driven has returned to where it was before the pandemic, but the risky driving behaviors that we saw during the pandemic -- specifically, driving at high speeds and distracted driving -- have persisted.
In terms of underwriting, we want to go beyond where most are now. Those using telematics have a customer drive for three to six months, then determine how good a driver the person is. We want insurers to have that data at the point of sale, not six months later.
How do you gather the data so you know what kind of driver I am when I show up on the doorstep of, say Allstate, to buy insurance?
Only about 16% of Americans are currently in a telematics program, but over 80% of Americans have a smartphone through which they're already sharing their location data. They’re doing it with everyday consumer apps for finding good gasoline prices, checking the weather, etc. So, Arity has forged relationships with these mobile app publishers, who gain permission from users to monitor their driving behaviors. We power some of the features that consumers can benefit from, such as crash detection that could lead to the deployment of emergency services or roadside assistance. In exchange, consumers consent to share driving data with us.
We have over 45 million active connections across the U.S., and insurers can tap into the database. When John Smith is shopping for auto insurance, they’ll ping us to learn about his driving behavior and use the data to price the insurance.
A decade ago, assessment of risk via telematics largely focused on speed and hard braking. Are there other behaviors you’re monitoring now, as well?
Those are core behaviors that are still highly predictive, but the technology has expanded. We can also monitor phone distraction as a predictor of risk. As you know, distracted driving is almost an epidemic in the U.S.
We're also looking at driving behaviors that are contextualized. By that, I mean it's not just about speeding – is this person driving over 80 mph? It’s, is this person speeding versus the posted speed limit?
We’re also looking at daily usage patterns. How frequently does somebody drive and for how long? What’s the driving environment? That's taking into account the types of roadways that an individual is driving on, which could even factor in dangerous intersections. Some roads and intersections are more dangerous than others.
So, you’re not just looking at me as the driver. You're looking at where I'm driving and when I'm driving and layering those risks on top of my behavior.
What kind of pushback do you get on what you're measuring?
People might say, I was driving, but I wasn't the one using my phone, so I shouldn't be getting dinged for phone distraction. Or, I wasn't the driver. I was the passenger. If that’s the case, the user can easily update that in the app
Another issue is the Big Brother effect. People may say, I just don't want my insurance company tracking me.
But, more and more with smartphones, people see this as an exchange. Am I willing to provide my data in exchange for the potential to save on auto insurance? Because if you think about it, driving behavior is an actual measure of driving risk. It's not a proxy, like a credit score or your age or your education. We did a survey back in 2021, and the majority of folks said, “Yeah, I would rather be priced and assessed based on my driving behavior and my driving record, versus who I am and where I live.” So, I think that Big Brother concern is becoming more and more the minority.
Do I have the opportunity to see what your driving data is, and do I get an opportunity to protest or correct it. I'm thinking of the credit bureaus, such as TransUnion, which these days have to let me know what my score is and protest items if I want to.
That’s a really good insight. Arity treats our ArityIQ product, which is pulled at point of underwriting, as a FCRA (Fair Credit Reporting Act) product.
So, it's a consumer report, similar to credit data, and Arity abides by FCRA rules, which include consumer disclosure. Individuals have the right to request their consumer disclosure which would include information that was shared with an insurer for quoting or underwriting purposes. They also have the option to dispute information if they believe it is inaccurate.
Where do you go from here?
A big thing is to continue to grow our database of connections. We're at over 45 million active connections right now. That represents just under 20% of the U.S. population. We want driving behavior on the majority of the population. So, we need to continue to grow our connections with mobile app publishers, but at the same time also continue to add connected car data from OEMs [original equipment manufacturers].
The other thing is that, right now, our product is based on providing an insurer with a score. That insurer uses that score to determine whether the customer shopping for insurance should get a discount. Our next evolution is providing what we call back attributes, the actual driving behaviors that go into the score. Based on our interactions and market research, that approach really resonates with, I'll say, the top 10 auto insurance carriers. As you know, all of them have their own telematics program, and they all have their own secret sauce, so they want the ingredients, those driving attributes. They’ll take those ingredients and score them themselves.
What’s next in terms of the analysis you’re trying to do?
We’re looking at other more predictive attributes. An example is what I'll call contextual braking. What were you doing when you hit the brakes hard? How fast were you traveling? Was it during the daytime or at night? Was it during a rush-hour period? Using more of these contextual attributes helps provide even more predictive lift in terms of risk assessment.