Property and casualty insurers are closer than they have ever been to pricing an individual policyholder’s risk rather than lumping similar homes or drivers into broad rating groups, according to the actuaries and economists who build the models carriers use to set premiums.
Industry leaders, including those at the American Property Casualty Insurance Association (APCIA) – the trade group representing roughly two-thirds of the US property-casualty market – say the data now exists to get close to pricing risk at something approaching the individual level, rather than relying on the broad, homogeneous categories that have defined rating for generations: wood-frame homes near a given fire hydrant, drivers under a certain age. The shift, driven by an explosion of new data sources entering carriers’ pricing models in 2026, promises sharper and arguably fairer premiums. It is also forcing a slower-moving negotiation: convincing state regulators that the data behind it will not be misused.
Judson Boomhower, a professor of economics at the University of California, San Diego, has studied how major US insurers price wildfire risk and found wide variation in how finely carriers segment their books of business. Some build highly granular models. Others price across far broader categories and correct for the resulting selection problems on the underwriting side instead. The difference, he said, comes down to cost as much as capability.
“It’s not at all free to have this very sophisticated and granular pricing schedule,” Boomhower said. “It seems like there are some players for whom it’s been worth it to make the big investment in this more sophisticated pricing, and there are other players for whom the right choice has been to hold off on making those investments.”
Regulatory approval adds a second layer of cost. In conversations with insurers, Boomhower said, the upfront expense of a more complicated pricing system is weighed against more than backend technology costs. It is also weighed against “the regulatory burden involved in getting approved to price in a particular way,” which he said “can be non-trivial” in states with more stringent review processes.
The economic case for individual pricing is straightforward: the closer a premium reflects one person’s actual risk, the stronger the price signal to invest in mitigation, whether that means replacing a roof or managing vegetation around a home. Boomhower argues that gain comes paired with a real cost.
“As you customize prices more and more to reflect one individual person’s level of risk, there’s a trade-off,” he said. “You get the price incentives closer and closer to perfect. The downside is that you erode the risk protection” that broad risk pools have historically provided.
Historically, Boomhower said, pooling created extensive cross-subsidization, with lower-risk policyholders effectively helping cover higher-risk ones. As pricing becomes more individualized, that cushion narrows. He calls the exposure this creates “classification risk”: even a premium that accurately reflects a homeowner’s true risk no longer also protects them from the financial consequences of having built or bought in a riskier location years before insurers had the data to see it that way. “There’s going to be really important winners and losers in that,” he said.
Peggy Brinkman, a principal actuary at Milliman who has built pricing models across auto and residential property lines for more than three decades, said the bigger near-term obstacle for many insurers is not methodology but data volume. New data sources tend to start out sparse and improve over time, not the other way around.
“When all the telematics programs first started out, data was very scarce,” Brinkman said. “It’s very difficult to get a viable sample size.” Coverage improved only as voluntary pilots scaled into mainstream programs, she said, and aerial and satellite imagery followed a similar arc, with patchy and outdated early coverage that has since improved.
Credit report data, by contrast, was a rare exception: a complete, ready-made dataset insurers could draw on almost from the start. Brinkman said a newer category, in-home Internet of Things sensors that monitor for water leaks or electrical faults, is still in that early, low-sample-size stage and is not yet usable at the scale needed to build credible pricing factors. Until newer data sources mature the way credit and telematics data did, the carriers able to price closest to the individual level will tend to be the ones with the longest head start collecting it, not necessarily the ones with the most sophisticated models.
Brinkman said the rules governing what data insurers can use vary not just by state but by line of business within the same state, which complicates any simple narrative about which states are most restrictive. “There are more restrictions in Florida property than there are in Florida auto,” she said. “There’s more restrictions in California auto than California property, generally.” States with greater exposure to costly perils, or more contested legal environments, tend to draw closer regulatory scrutiny of how insurers use data, she said.
That patchwork is precisely what the industry must navigate to build the kind of regulatory comfort that broader use of individual-level data will require. Colorado has built the clearest framework so far: a 2021 law on unfair discrimination in insurance, still being implemented in stages, requires insurers to test and document how they use external consumer data and information sources (ECDIS) – everything from credit-based scores to geographic and behavioral data – in algorithms and predictive models, and to demonstrate to regulators that the results do not disproportionately harm protected groups. Rulemaking extending the law to private passenger auto and health insurance was still active in 2025, and several other states are watching Colorado’s rollout as a possible template.
A newer layer of property-level risk-scoring technology, built on computer vision and geospatial imagery analysis, is also narrowing the gap between large carriers with in-house data science teams and smaller regional insurers that previously could not afford granular pricing models, making individual-level assessment more commercially accessible across the market. Whether it becomes standard industry practice will depend less on what the data can do and more on whether regulators, and the consumers those regulators answer to, are convinced it is being used fairly.
