PropertyLens Launches Its Own New and Improved Crime Risk Model

New model measures reported crime against the national rate for every U.S. neighborhood, and pinpoints high-crime areas far more accurately than previous models.

San Diego, October 9, 2026 — PropertyLens, the property intelligence platform for insurance carriers, inspection companies and homebuyers, today announced the PropertyLens Crime Risk Model, a proprietary crime model that now powers every crime score in its API and property reports. The model estimates reported crime per resident for every census block group in the 50 states and Washington, D.C., and replaces the third-party crime index PropertyLens previously licensed.

"We've always believed that our third-party crime model was the one in our lineup we could improve the most," said Bob Frady, CEO and co-founder of PropertyLens. "So we did what PropertyLens always does: we found newer, better data and built our own."

The previous index ranked neighborhoods against each other on a fixed 0–200 scale, based on a projection of FBI crime statistics. The PropertyLens method used millions of national and local data points to create a more predictive ratio, based on actual crimes. In the PropertyLens Crime model 100 is the national rate and 200 is twice the national rate. An A grade now means the same thing in every market, and high-risk grades are reserved for the places where crime is frequently recorded.

How it works

The model starts with FBI crime reports for 2018 through 2024 from 14,348 police jurisdictions covering 342 million residents. PropertyLens then trained a model on 5 million geocoded police incidents from 69 cities to learn how crime is distributed within a city, using 77 public census, housing, employment, business and land-use variables. No race, ethnicity, nationality or language data is used.

Each property receives a grade, rating, score, national percentile and plain-language description for total crime, personal crime, property crime and seven individual offenses. Results cover the property's neighborhood plus the half-mile and one-mile areas around it, and are refreshed every November after the FBI releases new data.

Tested against real police records

PropertyLens tested both models against recorded police incidents in 67 U.S. cities, scoring every city with a model that never saw that city's data. The new model's neighborhood rankings correlated 0.81 with recorded crime, compared with 0.51 for the previous model, and it was more accurate in 66 of the 67 cities.

In Denver, Baltimore and Austin, which were withheld from training entirely, the neighborhoods the new model rated riskiest had 2.9 to 3.8 times their city's average recorded crime. The highest-rated neighborhoods under the previous model had only 1.3 to 1.5 times the average.

What it means for customers

For insurers, the result is crime grades that separate genuinely high-crime neighborhoods from the safer areas around them. The share of U.S. residents in a D or F area falls from 22% to 10%, so underwriting rules keyed on crime flag the places that warrant attention rather than whole metro areas. Two new fields, personal crime and property crime, let carriers treat bodily crime and theft crime separately.

The PropertyLens Crime Risk Model is live now in the PropertyLens API and in all PropertyLens property reports. Existing API integrations continue to work without changes. It joins PropertyLens' other in-house models, including the Structure Fire Score and Water Quality Grades.