Why CurbFacts
Anyone can put a number on a house. We publish where every number came from, what we checked it against, and what happens if it's wrong — because a report you can't audit is just an opinion with a price tag.
The standard
Every fact in a CurbFacts Report names the record it came from and the date we pulled it. No source, no print — that's the rule the whole product runs on. The coverage counts and documented findings behind the reports are published on the research page.
When two public records disagree, the report prints the disagreement and names the document that settles it. A conflict between sources isn't an embarrassment to bury — for a buyer, it's often the most valuable line in the report.
The federal flood map is the authority — not a listing site's model of it. We query it first; if it doesn't answer, a registered mirror of the same federal data does, and the report names which source answered. Every feed we republish also has to reproduce a known answer against ground truth before we trust it with your address. Feeds that fail don't run.
Modeled numbers say "modeled." The tax-reset and true-cost math shows its work, so you — or your lender, or your agent — can check every step.
The Eyes Open Score uses the same published weights for every home in America. No demographic data appears in any layer or score. Ever.
If a fact is wrong, we re-verify it against the primary source within 48 hours. Confirmed errors are corrected for every holder of the report and logged — in the open, like everything else. Read the corrections log.
The method
The Eyes Open Score (0–100) answers one question: how much homework remains on this home? Higher means less remains. It is computed by one published method for every address, from seven layers, each with a fixed weight. Nothing demographic is ever an input.
| Layer | Weight | What it measures |
|---|---|---|
| Price vs. evidence | 20% | Asking price vs. an independent value estimate, same-street sold prices, and the home's own sale history |
| Flood & insurance exposure | 20% | The official FEMA flood zone at the exact address point |
| Safety | 15% | Crime level and trend, and sex-offender registry counts (counts and distance bands only — never names or photos) |
| Cost predictability | 15% | What the buyer's property tax will actually be after the sale, vs. what the seller pays today |
| Schools | 10% | Nearest public schools' published proficiency, ranked within their state |
| Noise & surroundings | 10% | Proximity to major roads, rail lines, and airports, computed from open map data |
| Development & change | 10% | Permits, rezonings, and large projects nearby, from published city/county open data |
The asking price is compared with an independent automated value estimate (RentCast AVM, dated in the report): asking at or below the estimate scores near the top; every 10% above it lowers the score on a published ladder. Same-street sold prices and the home's own history adjust it further — a quick relist at a large markup, or a falling price sequence, deducts points, and every deduction is itemized in the report.
Scored on the official map fact alone: high-risk zones (VE/V) score lowest, Zone AE/A next, then the remaining special-flood zones; moderate-risk shaded X sits higher; unshaded Zone X or no flood polygon at the point scores highest. A hazard boundary within about 500 feet of a point that is not itself in the floodplain still deducts — near the line is homework too.
Crime and registry evidence, blended 70/30. Crime uses published incident counts and rates (FBI county data, or a sourced ZIP-level compilation where one exists) compared against all areas at the same grain. The registry component uses residents-per-registrant: about 1 per 250 residents scores low; 1 per 2,000 or sparser scores high. The report always labels which grain answered — ZIP or county.
In states where taxes reset at sale, the buyer's tax is estimated from the property's own effective rate applied to the purchase value, printed as a range with the formula shown. The score falls as the buyer's tax rises above the seller's capped tax — a doubling is heavy homework; no reset (buyer ≈ seller) scores high. High effective tax rates deduct further.
Official data only: the nearest public elementary, middle, and high school (NCES), each school's published reading/math proficiency (EDFacts) converted to a percentile among schools at the same level in the same state, then averaged. Labeled "nearest," never "zoned," and the data year is printed — federal school-level proficiency currently publishes through 2020.
Proximity, computed from OpenStreetMap data — not a sound measurement. Published distance bands deduct for motorways, primary roads, heavy rail, and airports (an airport within a mile deducts most; within five miles, least). If the map data doesn't answer, the layer doesn't guess.
From published city/county open-data APIs only: active rezonings within about a mile deduct 12 each (18 within a half mile), large projects (over $1M or 50+ units) within a half mile deduct 10 each, ordinary permit volume is context only. Coverage is an explicit, growing list of jurisdictions — where no lawful open-data source exists, this layer reports no answer rather than inventing one.
If a layer genuinely cannot answer for an address, the score is computed from the layers that did, with their weights rescaled in proportion — and the report prints the score, which layers fed it, and their effective weights. Worked example: if development (10%) can't answer, price's effective weight becomes 20 ÷ 90 = 22.2%, and so on for each answered layer. Never a silent partial score, never a number adjusted to look right. A report where all seven layers answer carries the FULL coverage badge; otherwise it is labeled CORE and says exactly what is missing.
The proof
5377 Gene Circle, West Palm Beach. Zillow displayed minimal-risk Zone X. The official FEMA map put the address point in Zone AE — a Special Flood Hazard Area, with different insurance costs and lender requirements. The lot sits near the line — points ~130 ft away are Zone X — which is exactly why the report queries the point itself, not the neighborhood. Our report led with the conflict, because the conflict was the story. Read the sample report
354 Edmor Road, West Palm Beach. Sold October 19 for $1,000,000 — relisted the next day at $1,500,000. That price history is a public record; the listing just doesn't volunteer it. The report does, on page one.
The comparison
Built to sell you the home. One flood answer, from a model, shown once. No verdict, sources rarely cited — and they're paid by the sell side, not by you.
Built to rank in search. The same raw public records, dumped into tables — errors, $0 sale rows and all — with no checking, no verdict, and nobody's name on the result.
Built to be believed. The official records, cross-checked, with conflicts printed, a 0–100 verdict on top, and a source and date under every fact. Paid for by the reader — only.
Questions
Public records only — FEMA's official flood maps, county assessor and recorder files, state sex-offender registries, and published school and crime data. Every report cites the record behind each fact, so you never have to take our word for anything.
Tell us. We re-verify the fact against the primary source within 48 hours, and if we're wrong the correction is issued to every holder of that report and logged. Accuracy is the product — we can't afford to be casual about it.
Never. No report layer and no part of the Eyes Open Score uses race, ethnicity, religion, or any other demographic data. The report covers the property and its public records — the same report, the same way, for every address and every buyer.
They hand you the same raw public records we start from — unsorted, unchecked against each other, with data errors left in. We do the cross-checking, print the conflicts, and sign our name to a verdict.
Run the free preview. If the full record matters — and on the biggest purchase of your life it does — the full report is $29.