PF PFAS Decision Engine
Menu
Methodology

How the engine turns evidence into next actions without hiding the decision rules.

The project behaves like a typed editorial system. Sources are captured, normalized, mapped into deterministic routes, and rendered into the same action surfaces used by the public site and internal JSON.

Check your PFAS next step Read source policy
Verification ledger

Verification stays attached to the route.

Last verified 2026-03-20 / Methodology and trust layer

Verification details
Editorial owner
gabi Editorial Team
No external reviewer claim is made on this build.
Methodology owner
gabi Water Evidence Team
Read methodology for source handling and route logic.
Scope
Methodology and trust layer
Action-first, not safe or unsafe and not a health diagnosis.
Sources
243
Tracked source documents in the registry.
State routes
10
State guidance seeds currently in the engine.
Public-water dossiers
43
Seeded public-water systems with direct context.
Readiness
53
Routes currently marked ready for controlled expansion.
Pipeline

Evidence becomes a route through a typed pipeline.

The public interface is the final render step, not the first place where the logic shows up.

01

Raw capture

Official pages, PDFs, notices, and listing records are captured with URLs and retrieval dates.

02

Normalization

Facts are normalized into CSV or JSON with source ids, dates, and units.

03

Decision layer

Deterministic rules map source type, benchmark relation, and filter state into next actions.

04

Result rendering

SSR pages and JSON share the same typed result schema.

What the engine does
  • Routes public water and private well households differently.
  • Prefers direct official evidence over indirect hints.
  • Keeps certification, cost, and maintenance visible before a product feels like a fit.
  • Uses programmatic readiness checks before state or utility coverage is treated as expansion-ready.
What the engine does not do
  • It does not label a household safe or unsafe.
  • It does not turn ZIP hints into exposure claims.
  • It does not treat certification as proof of every current regulatory benchmark.
  • It does not use reviewer theater or fake expert voice to turn uncertainty into certainty.
Persona lanes

YMYL posture is handled as persona routing, not authority cosplay.

The engine does not pretend to be a clinician, regulator, or household inspector. It names the household posture, the evidence class, and the next responsible move.

Public-water starter

Read the utility record before the product layer.

The first move is CCR, utility PFAS page, or direct notice. Geography and shopping intent stay subordinate.

Private-well owner

Use state guidance and lab context before hardware certainty.

Testing, units, comparability, and state notes come first. A private well is not rendered as public-water compliance.

Result-holder

Unknown and not comparable are valid result states.

The engine keeps uncertainty visible and routes back to the missing evidence instead of turning it into a shopping cue.

Buyer-in-motion

Comparison opens only after the route is defendable.

Certification, annual cost, and cadence stay visible beside the route so the buying step cannot outrun the evidence.

Decision gates

Commercial unlock rules are part of the methodology, not a merchandising layer.

Keep comparison closed when

  • No direct utility record exists for a public-water route.
  • No well test exists for a private-well route.
  • The result is unknown or not comparable.
  • The user is really asking a whole-house question without a separate purpose case.

Open comparison only when

  • The evidence layer is direct enough to support a treatment discussion.
  • The route already narrowed to ingestion-focused point-of-use.
  • Claim scope, annual cost, and maintenance burden stay visible on the same screen.
  • Whole-house is still treated as a separate escalation review.