Methodology · rules v1.0 · updated 2026-09-11
How We Analyze Water
Every number on PurityMap traces back to a dated, published record. This is the reference that reports, rankings, and recommendations link to; it changes only with a version note.
Method boundary
A score is not a safety verdict.PurityMap summarizes published supplier evidence. It cannot certify the condition of a household tap.
Last method update 2026-09-11
1. From source to published intelligence
Source
Utility CCRs, state SDWIS records, and EPA UCMR5 analytical results.
Structure
Publisher, dates, sampling location, unit, qualifier, limit, and source stay attached.
Validate
Incomplete or contradictory evidence is retained as a visible limitation.
Assess
Usable measurements are compared only with their stated benchmark and finding rules.
Match
Actionable concerns are matched to verified treatment capabilities before offers exist.
Findings are compiled from utility Consumer Confidence Reports (CCRs), state Safe Drinking Water Information System (SDWIS) compliance records, and EPA UCMR5 analytical results. Each record keeps its source document, publisher, publication date, sampling date, sampling location, reporting limit, unit and result qualifier. Sources are listed at the bottom of every report and on the sources page.
2. Finding states
A finding state describes what the evidence supports. Missing, unusable, and unmonitored records remain distinct from a laboratory non-detect.
A measured value exceeds the cited federal, state or health guideline. Treated as a required capability by the recommendation engine.
Detected, but within the guideline. Lead and copper below the action level are treated as precautionary; everything else needs no treatment.
Analysed and not found above the laboratory reporting limit. Never rewritten as "safe"; the reporting limit is shown.
The supplier has not published a result for this substance. A missing measurement is never treated as a non-detect.
A result exists but cannot be interpreted (missing units, qualifier or date). No score contribution, no recommendation.
The record is contradictory or unverifiable. Shown so the gap is visible.
3. Purity score
Comparison summary, not certification
The purity score (0–100) summarises a supplier's published findings for comparison between reports. It is not a safety rating. It is computed only from findings with a usable measurement, so a supplier that monitors more substances is not penalised for transparency, and a supplier that has not published a result is not rewarded for silence. Health findings above guideline weigh most; aesthetic characteristics (hardness, taste) weigh least. The exact weights are versioned with the rules and shown on each ranking snapshot.
4. Treatment recommendations (rules v1.0)
Recommendations are generated, not written. The engine assesses every finding into a tier, then matches actionable concerns to products through verified certification claims:
- Required capability — a health finding above its guideline. A product qualifies only with a verified claim (for example NSF/ANSI 53 or 58) for that exact substance, on a standard the product carries.
- Precautionary — lead or copper detected below the action level. Shown as optional, because no level of lead is considered safe.
- Household preference — hardness and other aesthetic characteristics. Never presented as a health treatment.
- No treatment indicated — not detected, not monitored, unknown or stale findings. Listed so that gaps are visible; no product is attached.
Household preferences (renting, no drilling, installation type, maintenance tolerance) exclude products with a stated reason. Merchant offers are looked up after eligibility is decided and can only attach a link; commissions never add, remove or reorder products. When no eligible product covers a health finding, the page says so instead of hiding it.
5. Publication rules
Every report carries a publication status: draft, reviewed, published, withdrawn or insufficient evidence. Only published reports enter the sitemap. A contaminant page for a city is published only when at least one dated measurement backs it. Rankings are immutable snapshots with a metric, data cutoff, eligibility rules and an exclusion list, so a reviewer can reproduce them; an article cites the snapshot it was written from and links to newer ones.
6. Editorial process and AI assistance
Pages are prepared with AI assistance and reviewed by the PurityMap Editorial Team before publication. AI is used to draft explanations, summaries and FAQs from the structured findings; it is not used to invent measurements, sources, testing or reviewers. Product analysis is document-based unless a page explicitly describes hands-on testing. Corrections can be submitted from any page and are checked against the cited source.
7. Version history
| Version | Date | Change |
|---|---|---|
| rules 1.0 | 2026-09-11 | First published engine: tiers, verified-claim gating, preference exclusions, offer independence. |
Frequently asked questions
Is the purity score a safety rating?
No. It is a comparison summary of published findings. Safety questions belong to the specific finding, its guideline and your household circumstances.
Why is a substance shown as "not monitored" rather than left out?
Because a missing measurement is a fact about the evidence, not about the water. Hiding it would imply the water was tested and found clean.
Do affiliate commissions influence recommendations?
No. Eligibility is decided from findings and verified claims first; offers are attached afterwards and can only add a link. This is enforced in code and covered by automated tests.