Result first. Evidence second. No invented local price.

1. Decision input

The plan accepts a radon reading, a five-digit ZIP, and whether the user is living in, buying, or selling the home. An invalid reading is never coerced into a valid band. The plan does not claim the test protocol was valid because duration, device, placement, weather, and closed-house conditions are not collected.

2. Decision rule

Versioned logic assigns the reading to an action band and returns no more than three next steps. EPA guidance is the national baseline. Transaction context changes wording and sequence, not the underlying health threshold.

3. Location evidence

ZIP is mapped to a primary county and can be ambiguous near boundaries. Official measured summaries are preferred; official tiers and EPA zones are context only. A county average or zone never substitutes for the entered home result.

4. Private snapshot

A share stores the generated answer with schema, decision, and data versions for 30 days. The raw random token is not stored; only its SHA-256 hash is. Shared pages are read-only, noindex, private/no-store, and omit the ZIP from the rendered page.

5. Search inclusion

Mass county publication is retired. Five same-path county pages remain as preservation canaries, not proven winners. Other generated county pages and all modeled local-cost surfaces are excluded or returned as gone in production.

6. Known limits

• No result can be validated without test-protocol details.
• Public datasets vary by period, method, completeness, and geography.
• Rules and credential requirements can change; verify with the official state program.
• The software is informational and is not medical, legal, engineering, measurement, or mitigation advice.