Restoration Intel. Operator brief. For the restorer who has to sound precise to an adjuster, and for the adjuster who has to hear what the week actually did.
A cat model does not price your Tuesday. It estimates the chance a carrier’s book loses more than a given amount in a year. The street you are standing on is one address, one peril, and one week. Both numbers are real. They are not the same number.
What the model is actually doing
The carrier-side explanation lives on Risk Coverage Hub: the catastrophe modeling guide. Four modules. Hazard is the event set, the storms the model is willing to imagine. Vulnerability is how much damage those events do to a building of a given type. Exposure is the book, which roofs and which limits are in the portfolio. Financial is deductibles, limits, and the reinsurance that sits above the claim.
What comes out is a year-shaped answer. Average annual loss. The chance the year exceeds a threshold, which is not the same thing as the chance one event does. A probable maximum loss. No single vendor print is the loss. The names people use are Moody’s RMS, Verisk, and CoreLogic. A reinsurer checks those prints against its own view of risk. You should not treat one of them as the number on your invoice.
What this week is actually doing
Your side of the same weather is the storm desk. this morning’s storm pulse is a Moderate flood on the Southwest, with the severe map quiet on purpose. Sedgwick’s 2026 report, written up here as the 10-day gap between billion-dollar disasters, put the average gap between billion-dollar U.S. weather disasters at 10 days in 2025. In the 1980s it was 82. Labor and equipment do not reset between those events. That is an operating fact. It is not a model output.
The model can be right about the year and still tell you nothing about the ZIP you are in this morning. The pulse can be right about the ZIP and still tell a carrier nothing about whether the year is inside the loss they bought reinsurance for. The argument starts when someone uses one of those sentences to answer the other question.
How to say it to an adjuster
Do not open with a vendor’s name as if their print settles the claim. Ask which view of risk the file is being held to, and then describe the week in the same nouns the model uses, pointed at the address.
Hazard. Name the peril and the cluster. Hail last night and flood this morning are two events inside a short window, not one dramatic storm with a name. The 10-day gap between billion-dollar disasters is the industry evidence that the window has gotten short. Your photos are the evidence at the address.
Vulnerability. The model uses a damage ratio for a class of building. You have the meter map, the category of water, and what the assembly actually did. A Class 3 or a sewage loss is not the average house in the event set.
Exposure. The model sees a portfolio. You see one policy, one deductible, one limit. Say that out loud. The book can be priced for a bad year while this file is still a single location.
Severity you can see. Cycle time is the part of the loss the event set does not itemize. If the crew and the equipment cannot reset because the next event is already on the desk, that delay belongs in the file as dates, not as a speech about climate.
What not to do with the model
Do not tell a homeowner the carrier’s model proves the claim should pay more. The model is not a coverage opinion and it is not your scope. Do not pick a fight with a vendor you have not read. The useful sentence is smaller. The year can look normal in a model while this week is not normal on the street. Here is the week. Here is the address. Here is what the water did.
Operator brief. The carrier-side definitions are on the catastrophe modeling guide, last verified 9 September 2026. This page does not produce a model number, and it is not a coverage position.