Territorial ratemaking – or using geographic areas as proxies for determining driving risk – hasn’t kept up with how people actually drive. Regulatory constraints and sparse data make it hard for insurers to assess risk and set accurate and fair rates with confidence.
Todd Lehmann, Chief Risk Officer and Chief Actuary at Quincy Mutual, knows the challenge firsthand. With a career that spans predictive modeling, underwriting, product management, and pricing at companies like Allstate and OneBeacon, Todd now leads risk strategy at Quincy – a New England-based mutual with over $570M in direct plus assumed written premium and almost $1.7B in surplus.
He sat down with Arity’s Henry Kowal, Product Director. He shared how Quincy is pushing past the limits of traditional territory models, as one of the early adopters of Geosight – Arity’s new driving behavior dataset.
Geosight is the anonymized and aggregated output of Arity mobility data collected daily from a wide range of sources — including mobile apps and in-car systems — with clear user consent. Arity collects this data only from individuals who opt in, ensuring both scale and privacy.
Quincy Mutual evaluated Arity’s Geosight dataset, updated their pricing models, and filed new territorial relativities – all in a span of about four months.
This interview has been edited for length and clarity.
I put them in three different buckets: regulatory, data, and our distributional footprint. In terms of the regulatory part of it, in New England, there are a number of territorial rating restrictions. Connecticut and, similarly, Massachusetts, will not allow towns to be divided into zip codes or census block groups for the purposes of insurance pricing. There are other regulatory restrictions on variables that you can or can’t use. Connecticut regulations require subsidies for certain towns like Hartford and Bridgeport.
In terms of data, outside of Massachusetts, we don’t have significant market share, so data sparsity can be an issue. Even in Massachusetts, there might not be enough data, as our market share generally ranges from 1% to 5% from town to town.
And finally, in terms of footprint, Quincy sells through independent agents. While it is debatable, we believe our agents place better business with us because we are local to New England, we understand the needs of the customers in the region, and we treat their best customers fairly. An agent that underwrites for us in a particular market can distort our territorial pricing for that town or local area if we aren’t paying attention.
We use our own premium and loss data, of course. We also look at competitor rate relativities to see where we are in the marketplace. We use third-party data as well. In Massachusetts, we use data available through the Registry of Motor Vehicles as well as census data. In addition to Geosight data, we pull in vehicle data, industry data, and demographic information such as population density that helps us understand the population of all drivers and vehicles better, particularly when we know we write policies filtered to our target preferred segment of the personal auto business.
We underwrite through independent agents, which might skew the population based on the volumes of policies that agents write. We tend not to underwrite high-risk drivers, so we might not have a full picture of rates by territory or class.
Also, the recency of the data – many data sources use pre-pandemic and pandemic data, but the volatile environment post-pandemic makes it difficult to set accurate rates.
Geosight gives us visibility we simply didn’t have before.
We think of providers as partners, not vendors. There needs to be a partnership. Price matters, but so does responsiveness, depth, and the breadth of the data and how it has been captured and normalized. Data quality is key.
Geosight gives us data that covers driving behaviors across a wide sample of the driving population. And it’s aggregated and anonymized, not capturing the data of individual drivers, so it’s privacy safe. The time component is useful as well. You can look at trends and changes over time at a territory level. It’s predictive.
It’s more recent. We’re getting a real-time, monthly view of what’s changing. It gives us another source to validate or assess our own data. We can use it as a reasonability check and it could be a complement of credibility to our own data.
When an agent, underwriter, product analyst, or actuary wants to understand shifts in driving behavior at the ZIP code level, there’s no better tool.
Driving behavior differences at the territorial level were fascinating. Seeing how behaviors changed over time – before, during, and after the pandemic – made us think about how this data could supplement our own loss experience. It brings in drivers and behaviors we don’t currently see in our book but are in our territory.
I mentioned the recency of the data. It comes out every month, that’s about as recent as you can get. It also provides another source to validate or assess our data. If we are seeing trends in frequency in a particular area or town, we can look and see if that’s showing up in our data. We can use it as a complement of credibility or as a reasonability check.
Geosight gives us visibility we simply didn’t have before. When an agent, underwriter, product analyst, or actuary wants to understand shifts in driving behavior at the ZIP code level, there’s no better tool.
Even if we had a telematics program, it would only reflect our customers. Instead, Geosight goes broader. It cuts across niche, segment, and tier to show the true magnitude of change in a territory.
Book a demo, and see how driving behavior data can help you respond faster, price smarter, and uncover risk you may be missing.