Methodology
Where every number comes from.
Somewhere never shows a figure without a source and a date, and it never collapses the data into a score or ranking. Estimates are labeled as estimates, with their assumptions inspectable. When a source is unavailable, the section says so — we don't substitute a plausible-looking number.
Sources & update cadence
| Data | Source | Refresh |
|---|---|---|
| Geocoding | Google Geocoding API | Cached permanently per address |
| Travel times | Google Routes (typical weekday, 8:30 AM) | ~90 days |
| Property records & value | RentCast (AVM estimate or last sale) | ~14 days |
| Home price areas | Redfin Data Center (median sale $/ft²) | Monthly |
| Places nearby | Overture Maps | Quarterly |
| Schools | NCES locations — nearest, not assignment | Annual |
| Flood zones | FEMA National Flood Hazard Layer | On FEMA publication |
| Climate | NOAA 1991–2020 station normals | Static until next normals |
| Electricity rates | EIA state residential averages | ~90 days |
| Job hubs | Census LEHD LODES (jobs counted at the workplace) | Annual, per state |
| Property tax rate | Census ACS 5-year — estimated, see below | Annual |
Estimates
Monthly cost projections combine the property record with your adjustable mortgage assumptions (default 20% down, 6.5%, 30 years). Electricity models usage from home size and local heat; insurance is a flat share of price, increased in high-risk flood zones. Property tax shown monthly is exactly the annual figure ÷ 12 — one source, reconciled. Travel times are modeled typical-weekday averages shown as ranges, not live traffic.
The citywide effective property-tax rate is an estimate, not a published assessor rate: no agency publishes a national effective-rate feed, so we divide the county's median real estate taxes paid by its median home value (Census ACS tables B25103 and B25077). It is sound for comparing counties and nothing finer — assessed value trails market value, exemptions apply unevenly, and rates vary by district within a county. Your actual bill will differ.
Job hubs are derived, not published. Census LODES counts jobs at the workplace by census block; we snap those blocks to a 1 km grid and cluster them into employment centers, positioning each hub at its job-weighted center. Hub boundaries are ours, so the job totals describe a district, not a city-defined neighborhood.
What we deliberately don't do
No composite scores, no rankings, no "best city." No neighborhood quality or safety scores derived from demographics. Regional exposure layers are not parcel-level risk — the address report carries the parcel-level FEMA zone with a "verify before purchase" caveat.