Methodology & Data Sources

How the site produces results

Most tools read saved datasets bundled with the site and run deterministic calculations. They do not ask an AI language model to produce an answer. Address lookup and parts of the walkability map make live mapping-service requests. Comparison introductions are generated from templates; editorial articles are stored separately.

The optional move-planner chat uses OpenAI GPT-5.4 mini to extract structured details from your message. A server-side validator checks those fields; the existing calculators then generate estimates from saved data. The model does not supply prices, execute code, browse the web or determine a household’s school eligibility.

A source name is not a guarantee that every number has been independently validated. Some inputs are imported datasets; others are saved tables or model assumptions whose original citations need further verification. Estimates should be checked against current local information before making a housing or financial decision.

Equivalent salary and monthly difference

The comparison reads the saved overall cost index for each city, applies any city fallback adjustment, then uses this formula:

The 60% spending share is a fixed modeling assumption, not a measurement of your budget. The monthly difference is (current salary − equivalent salary) ÷ 12. For example, an index ratio of 0.80 with a $75,000 salary produces a $66,000 equivalent salary and a $750 monthly difference. This is not guaranteed cash savings or a take-home-pay calculation.

Category estimates use fixed national spending amounts labeled as BLS-based in the code, scaled by salary and saved category indexes. The original derivation of every input has not been verified. The calculator uses the stored overall index for equivalent salary; it does not rebuild that index from the displayed category amounts on each request.

Cities without their own cost table

A city can use a mapped data city or metro primary city. For cities with a data-city mapping, the model adjusts the borrowed values using distance, population, and regional price parity. It prefers an available metro parity, then state parity, then an index of 100.

• Up to 25 miles: no regional-price blending, although the population adjustment can still apply.
• From 25 to 100 miles: the blend increases linearly toward the regional-price target.
• Above 100 miles: a separate branch starts near 70% regional weighting and reaches 100% at 400 miles. This creates a discontinuity just beyond 100 miles; estimates near that boundary deserve caution.
• A smaller-population adjustment is bounded at 65%; the combined cost adjustment has a 55% floor. These are model choices, not directly observed local price discounts.

The current comparison also applies its adjustment factor to rent values, including ZIP-derived rents. A displayed adjusted estimate can therefore differ from the underlying HUD value. We have not established independent validation of these heuristics and do not attribute them to a calibrated USDA model.

Dataset families and coverage

Neighborhood estimates and walkability limits

The older neighborhood builder estimates restaurant, park, and hospital counts from population. Its walkability value is derived from population density, and its safety score transforms a saved crime-rate table. These unsupported neighborhood scores and counts were removed from the public guides and client portal on September 30, 2026. The guides now provide editorial place names, map links and an address research checklist. The separate OpenStreetMap road-attribute snapshot remains an explicitly labeled model.

Read the walkability methodology and snapshot limitations for the city-score formula, missing-tag behavior, and geographic validation limitations.

Freshness and corrections

These are saved snapshots, not a continuously refreshed data feed. Source years and coverage differ. Having an import script does not mean it runs automatically or that the latest upstream release has been loaded. Missing observations and fallback assumptions can materially change a result.

For questions or corrections, use the contact link on our About page. Include the page, city, and value that appears incorrect.