Legal
Methodology
Last updated 26 July 2026
This page describes exactly how a HealthDeed Score is produced. It is written to be checkable: if you disagree with a weighting or a rule, you should be able to see precisely what it is and argue with it.
The core rule: only property-specific data is scored
Every indicator we collect is classified by the resolution of the data behind it. That classification decides both how the figure is displayed and whether it counts toward the score at all.
- Measured — point or parcel-level data for this address, such as a measured distance to a mapped EPA facility, or the flood zone the parcel sits in. Shown with its value and units.
- Modeled — real data at census-tract, water-system, or county level. Shown with its value and its resolution stated plainly.
- Contextual — a state-level baseline table with no per-address signal. Shown as a qualitative band only, never as a number, and excluded from the score entirely.
- Pending — requires an on-site inspection, so no public record can answer it.
The reason for excluding contextual indicators is simple: a state baseline is identical for every address in that state, so it cannot distinguish your house from any other, and including it would dilute the indicators that can. Those indicators still appear on the report, under “State & regional context”, because they are useful background — they just do not move the number.
A category is scored only if it owns at least one measured or modeled indicator. If none does, the category reads “Not scored”. If no category qualifies, no score is produced and the report says so rather than inventing one.
Category weights
The composite is a weighted average of the categories that could be scored, renormalised across them. Weights are:
- Water Quality — 20%
- Air Quality — 20%
- Soil & Contamination — 18%
- Natural Hazards — 16%
- Community Health Context — 16%
- Radon — 10%
Water and air carry the most weight because they are the exposure routes with the clearest evidence linking them to health outcomes, and the ones a buyer can least easily observe. Radon is deliberately modest at 10%: the public data is a county or state classification rather than a measurement, so it should be visible without being able to swing the result.
Sources and their true resolution
| Input | Source | Resolution |
|---|---|---|
| Address geocoding | U.S. Census Geocoder, OpenStreetMap Nominatim | Parcel, or ZIP centroid (disclosed) |
| Industrial and Superfund proximity | EPA TRI, Superfund NPL | Measured distance to mapped facilities |
| Drinking water compliance | EPA SDWA via ECHO | County water systems |
| Flood zone | FEMA National Flood Hazard Layer | Parcel (point in polygon) |
| Air quality | EPA AirNow | Nearest monitor, distance disclosed |
| Local health outcomes | CDC PLACES | Census tract |
| Radon | EPA Map of Radon Zones | County or state classification |
How proximity is weighted
Point hazards are distance-weighted, never counted as a simple in-or-out radius. Facility burden is derived from the distance to the nearest facility in stepped bands (under 0.25 miles, 0.5, 0.8, 1.2, then decaying), combined with a log-scaled count of facilities within 1.5 miles.
A sparse count is capped: one nearby reporting facility can produce at most a moderate burden figure, and a high one requires an actual cluster. This exists because the public data records that a facility reports to the EPA, not how much it emits — without release volumes, proximity alone cannot distinguish a scrap dealer from a refinery. Adding release volumes is planned, and until then the cap is the honest compromise.
Missing data
When a source cannot be reached, the report lists it as unavailable rather than silently substituting something. The confidence interval shown beside the score widens to reflect what was missing. We would rather show you a gap than a number we cannot stand behind.
Reproducibility
Scoring is deterministic: the same address scored against the same data snapshot always produces the same result. Every report is stored with the raw and normalised data it was built from, plus retrieval timestamps. A purchased report renders from its stored snapshot and does not change unless you choose to refresh it.
Known limitations
- Facility burden lacks release-volume weighting, as described above. This is the most significant open item.
- Straight-line distance ignores barriers. A facility across a wide river is treated as nearer than it functionally is.
- Drinking water compliance is county-level. Water systems do not map neatly to parcels.
- Radon zone data is a county or state classification, not a measurement. Only an in-home test tells you what your house is actually at.
We publish these because a methodology that lists no weaknesses is not a methodology, it is marketing.