How this is modeled
Real engine, modeled carrier inputs, and a clear line between them
CoverGuard is deliberately a working simulation of the carrier market while a live carrier feed is being integrated. That is stated throughout the product rather than buried, and this is the page behind the statement: what is measured, what is modeled, and what a modeled estimate is not.
Measured
The hazard layer
Flood, fire, wind, earthquake and crime exposure are read from public federal and state datasets — among them FEMA’s National Flood Hazard Layer and National Risk Index, USGS elevation, seismic design and fault data, NOAA storm and surge records, Cal Fire’s severity zones, the USFS wildland-urban interface, and FBI and Census statistics. These are real lookups against real sources.
Coverage varies by address and by source availability, so provenance is recorded per response rather than promised uniformly.
Computed
The scoring engine
The CoverGuard Score, its bands and the peril breakdown are computed by a deterministic engine from those measured inputs. The same address with the same data produces the same score. Nothing about the score is generated by a language model.
The CoverGuard Advisor is AI-assisted: it uses a language model to explain what the engine produced and recommend next steps, so its explanations can be wrong — but it never decides the number.
Modeled
The carrier layer
Which carriers appear (the ones licensed in the state), the appetite and writing status attached to each (derived from the property’s risk scores and the modeled market condition for that area), and the premium modifiers (a fixed figure per carrier) are modeled. They are not a live feed, and they are not invented per request either: the derivation is deterministic, so the same inputs return the same slate every time. Where a sourced record covers a carrier, its provenance comes from that record, and so does its writing status when the record carries one. The slate’s data basis then reads mixed, or live if every carrier on it is sourced.
There is no randomness in this path. A modeled slate is a reproducible consequence of the risk profile, not a guess that changes when you reload.
How the platform admits what it is
The disclosure is not a single banner. It is carried in the data itself, so a surface cannot accidentally present a modeled number as a sourced one.
The carrier slate’s data basis
A carrier slate carries its own provenance — SIMULATED, MIXED or LIVE — so a surface can state what it is standing on rather than leaving you to assume.
Appetite source & confidence
When a signal is modeled, the platform forces it to read as inferred, with low confidence and a timestamp from the moment it was modeled. It will not present a plausible-looking feed age that never happened.
The per-report line
The “Modeled estimate” note on a report is not boilerplate. It is the same disclosure as this page, attached to the specific number you are reading.
What changes when the live feed lands
The platform has a single switch for its data posture. When it flips to live, the Modeled data ribbon disappears and responses begin reporting live or mixed provenance instead of modeled. The engine, the scores and the perils do not change — they were never the simulated part.
Platform posture and per-response provenance are separate on purpose: a live platform can still return a modeled slate for a property with no connected feed, and it will say so.
What a modeled estimate is not
- Not a quote. No carrier has priced your property. Request a quote and CoverGuard’s brokerage team, which works with licensed insurance partners, takes it to the market.
- Not a binder. Nothing here places or confirms coverage.
- Not a promise about a named carrier. A modeled slate describes what the market looks like for a risk profile, not a decision any insurer has made.
Use it the way you would a comparable: to decide whether a property is worth pursuing, and what to ask before you bid.
Reviewing CoverGuard for diligence or procurement?
This page covers where the numbers come from. Security, privacy, accessibility and Advisor governance are on the Trust Center, and questions about a specific figure are worth asking us directly.
From the CoverGuard Record to a quote request — without the surprises
With carriers exiting major markets, insurance is killing deals at closing. CoverGuard reads the CoverGuard Record, models which carriers are likely to write the risk and what cover should cost, and sends your quote request to CoverGuard’s brokerage team — before anyone signs on the dotted line.
CoverGuard is in Private Preview. Ask for an invitation and a member of our team will review it — usually within one business day. Available in all 50 states.