How PropertyGenome scores neighborhoods, properties, and personas
Every score on the platform decomposes into named inputs with explicit weights and named provider sources. When a provider is missing for a parcel, the result reports that gap rather than hiding it behind a synthetic value.
How the 1–5 star Neighborhood Score is built
The Neighborhood Score is a 0–100 composite of eight inputs. Each input is bounded to [0, 100] before weighting. Inverted inputs (risk metrics) have higher = better applied. The final value maps to a star band.
| Input | Weight | Provider |
|---|---|---|
| School Quality | 30% | NCES + GreatSchools (district-level + nearest-school distance) |
| Walkability | 20% | OpenStreetMap (sidewalks, points of interest, intersection density) |
| Environmental Quality | 15% | EPA EJScreen, FEMA NFHL, NOAA storm history |
| Neighborhood Density | 10% | U.S. Census ACS demographics + parcel density |
| Inspection Risk (inverted) | 10% | County GIS code-enforcement inspections |
| Vacancy Risk (inverted) | 5% | Census ACS + listing-status duration |
| Violation Risk (inverted) | 5% | EPA ECHO + county code violations |
| Flood Risk (inverted) | 5% | FEMA National Flood Hazard Layer |
How match scores re-weight the same signals per persona
Every persona uses the same underlying signals (safety, schools, walkability, recreation, commute, rent strength, buyer demand, vacancy risk inverted) but applies a different weight vector. Pick a persona that matches who you're underwriting for and the same address can score very differently.
What data backs each layer
PropertyGenome blends federal, county, MLS, and licensed third-party feeds. Coverage varies by county — every report includes a data-audit panel that lists exactly which providers were used, which failed, and which signals fell back to synthetic values.
What you should always see in a report
- The score itself (0–100) and the band/stars it maps to.
- The named contributing signals and the weight applied to each.
- The provider that supplied each signal, or "coverage pending" when none did.
- A data-audit row counting real signals vs. synthetic fallbacks.
- The persona used and the alternative personas you could re-run.
- Last-updated stamp on the underlying provider feed.
Run a snapshot
Try the methodology on any address. The Neighborhood Intelligence page surfaces every input, weight, and provider for a single parcel.