USAA's Patent Turns Storm Density Into a Coded Severity Score

The U.S. Patent and Trademark Office granted USAA's US 12,682,402 on July 14, 2026: a method that draws a boundary around a weather event, profiles how densely insured properties sit inside it, and assigns a severity code from sensor data before a policyholder files a single loss report.

Read at claim level rather than as one line in a monthly tally, it is not a customer-service upgrade. It is a severity model that produces a numeric loss estimate before the first FNOL call (July grant roundup).

Key Takeaways

Three inputs before any output: a bounded geographical area affected by a weather event, a customer density profile inside it, and sensor data from that area. The density step converts storm geometry into a book-specific signal.
A six-level severity code routes the work, with independent adjusters at the bottom, preferred-provider networks in the middle and staff adjusters at the top, assigned before the claim is reported.
Pre-opening the loss report moves the reported date earlier without changing the loss. A triangle fitted to historical FNOL lag reads that acceleration as loss-cost improvement unless the reporting pattern is refitted.
Roughly 25% of claim adjusters are expected to retire by the end of 2027, against billion-dollar U.S. weather disasters now arriving about every 10 days versus roughly every 82 days in the 1980s.

The Claim: Boundary, Density, Sensor Code

Claim 1, filed October 1, 2024 as the latest link in a continuation chain reaching back to a March 2013 application and classified in the insurance data-processing art unit G06Q 40/08, recites three inputs before it recites an output. The system identifies "a bounded geographical area affected by a weather event," generates "a customer density profile in the bounded geographical area," and collects "sensor data from one or more sensors located in the bounded geographical area" (USPTO Official Gazette, July 14, 2026).

Those three feed a severity code, which in turn drives a prediction of resource utilization: which adjusters, which contractors, which materials, and where they go first.

The density step is the one doing the analytical work. A storm boundary alone tells a carrier where wind or water crossed a map. A density profile tells the carrier where its own book is thickest inside that boundary: the blocks where a hundred claims will cluster, against the ten arriving from a sparsely insured exurb on the same storm's edge.

Severity code	Estimated loss tier	Adjuster routing (specification examples)
1-2	Minor, routine damage	Independent adjuster
3-4	Moderate, contractor-scope repair	Preferred-provider network
5-6	Severe, total-loss or complex structural	Staff adjuster

This is not the industry's catastrophe threshold wearing a new name. Verisk's Property Claim Services unit designates a U.S. event a catastrophe at more than $25 million of expected insured property damage affecting a significant number of insureds and insurers, decided only after enough carriers report enough data. The claimed method runs inside one carrier's book while the event is still unfolding and never needs a PCS number. A hailstorm well under $25 million can trip a full severity-coding response if that carrier's policies happen to cluster in its path.

Staging Before FNOL Splits Into Two Effects

The back half of the claim opens loss reports for selected customers before those customers call, then schedules adjusters, contractors and materials against the predicted need. That is a direct intervention in loss adjustment expense. Pre-staged contractor capacity and pre-assigned adjuster skill cut the re-inspection trips and cross-referral churn that inflate ALAE on catastrophe files.

One effect is a genuine severity gain. A roof breached by wind and left uncovered for days accumulates water intrusion and mold exposure a same-day tarp prevents. That gain is worth more per claim than it was three years ago: all-peril homeowners claim severity hit an all-time high in 2025, up 25.9% year over year and 93.2% above 2019 levels, while all-peril frequency fell 23.8% (LexisNexis Risk Solutions, July 2026).

The second effect is pure timing, and on the diagonal it looks identical. Take a book whose history shows 60% of ultimate claim counts reported within 30 days of landfall, a lag pattern baked into the selected age-to-age factors. If pre-FNOL triage pulls that 60% mark to 10 days in the densest, highest-code neighborhoods, the observed 30-day count on the new accident quarter runs well above the historical benchmark even with ultimate claim count and ultimate severity per claim unchanged. Applying the old factors to that diagonal projects too high an ultimate.

The correction is not a severity trend selection. It is a reporting-pattern curve refitted to the post-adoption diagonal and kept separate from any severity adjustment layered on top. The effect is larger at USAA than at a geographically diffuse carrier, because membership eligibility runs through military service and the book clusters around installations sitting on hurricane and wildfire corridors. USAA responded to 62 catastrophes in 2025, paying nearly $5 billion at an average of nine days per catastrophe claim.

What a Pre-Opened File Does to the Claim Count

The unallocated expense side is where the mechanism complicates its own measurement. The two workhorse ULAE reserving methods, the paid-to-paid method and the claim-count-based Johnson method, both depend on a stable relationship between claims department expense and a claim-count or paid-loss base (Allen and Mango, Casualty Actuarial Society, Fall 1999).

A method that opens loss reports for "selected customers" ahead of any customer-initiated report changes what a claim count measures. A file opened for a property inside a dense storm footprint that turns out to carry no covered damage either inflates the count the Johnson method divides by, or it is filtered out through a definitional change to what counts as an open claim. Either way the ULAE ratio moves for reasons that have nothing to do with the cost of handling claims.

That is what makes the improvement hard to bank. The severity gain and the timing gain both surface as a faster, stronger first diagonal, and the claim-count base that would corroborate the expense saving is the same base the mechanism redefines.

The adjuster shortage supplies the pressure to deploy it anyway: "2026 is shaping up to be one of the most complex catastrophe seasons that carriers have ever faced, and the response models most carriers have in place aren't built for this new reality," said David Armstrong of Sedgwick (Claims Journal, June 2026), in the same reporting that counted 23 billion-dollar events in 2025 and put non-hurricane perils at 99% of that year's insured catastrophe losses.

USAA's Patent Turns Storm Density Into a Coded Severity Score

Key Takeaways

The Claim: Boundary, Density, Sensor Code

Staging Before FNOL Splits Into Two Effects

What a Pre-Opened File Does to the Claim Count

Further Reading

Sources