Skip to content
WaterQualityLens

Methodology

How we turn a street address into a defensible answer

WaterQualityLens is an auditor, not an advocate. Every result is produced by the same deterministic pipeline: resolve the water system, read its public record, and match only hardware that is independently certified for the specific hazards found. This page documents each step and the confidence we attach to it.

Point-in-polygon resolutionFull data provenanceConstrained set-cover matching

01Address resolution

Resolving your address to a water system

The supplier that serves your tap is a geographic fact, not a ZIP code. We geocode the address and test it against real utility service-area geometry using the three-tier TEMM approach — Tiered Explicit, Match, Model.

First we attempt an explicit boundary match: a point-in-polygon test that checks whether the geocoded coordinate falls inside a published EPA service-area polygon. When it does, the serving public water system (PWS) is known with certainty. When no polygon covers the point, we fall back to a matched proxy — the most probable system inferred from TEMM layers and corroborating identifiers — and, only as a last resort, to a modeled radius that infers the likely supplier from nearby systems.

Each outcome carries an explicit confidence score, and that score is surfaced in the report rather than hidden. We would rather tell you a match is uncertain than present a precise-looking answer we cannot stand behind.

  1. Tier 1

    Explicit Boundary

    1.00confidence

    Your address falls inside a published EPA service-area boundary — a point-in-polygon hit against a real utility footprint.

    In the report: The report is presented at full confidence, with no mapping caveat.

  2. Tier 2

    Matched Proxy

    0.85confidence

    No explicit polygon exists, so the address is matched to the most probable system using TEMM proxy layers and corroborating identifiers.

    In the report: The report is shown with a note to confirm the system name on your water bill.

  3. Tier 3

    Modeled Radius

    0.50confidence

    The serving utility is inferred from a modeled search radius when no boundary or proxy resolves — the least certain result.

    In the report: An amber banner recommends a physical lab test and verifying the utility on your bill before acting.

A Tier 3 modeled match is the one case where the map itself is uncertain. Those reports open with an amber banner recommending a physical lab test and asking you to confirm the utility name printed on your water bill.

02Data sourcing

Where the contaminant data comes from

We do not test water or generate our own measurements. We faithfully report what your utility has already submitted to the federal government — with its origin attached to every value.

Occurrence data is drawn from the EPA Safe Drinking Water Information System (SDWIS) through the Envirofacts REST API — a US Government public-domain source. We ingest updates on a weekly cadence so newly published sampling results and violation records appear without long lags.

Provenance is a first-class requirement, not a footnote. Every contaminant value in a report carries the four facts you need to judge it for yourself:

  • Sample date — when the measurement was taken.
  • Testing agency — who reported it.
  • Unit — the measured concentration and its units.
  • Legal MCL — the federal maximum contaminant level for context.

A number without its source is not information we are willing to show. Where a utility has reported nothing for a contaminant, we say so plainly rather than implying absence is proof of safety.

Ingestion at a glance

Primary source
EPA SDWIS
Access
Envirofacts REST API
Refresh cadence
Weekly
Per-value provenance
Always shown

Utility data is aggregated at the water-system level. It cannot see the plumbing inside your home — a legacy lead service line, for example. Where that limitation matters, we say so and point to a point-of-use lab test.

03Filter matching

Matching filters to hazards, honestly

Recommending hardware is a coverage problem, not a marketing one. We solve a constrained set-cover: find the smallest set of genuinely certified products that addresses every health hazard in your report — and eliminate anything that only appears to help.

The constraints are strict, and they are what make the answer trustworthy:

  • Only NSF/ANSI 53, 58, and 401 certifications can satisfy a health hazard. A product must carry a relevant health-effects certification to be eligible at all.
  • NSF 42, NSF 372, and the retired P473 are isolated from hazard matching. Aesthetic, materials, and obsolete certifications never count toward removing a contaminant.
  • Nitrate is reverse-osmosis only. Carbon-based pitcher, faucet, and under-sink filters cannot remove it; only NSF 58 (or specific ion-exchange) systems qualify.
  • Partial matches are eliminated. A product that covers some but not all of your hazards is not offered as if it were a complete solution.

The result is a short, defensible set of options in which the recommendation follows the data. Commercial considerations never reorder scientific fit; a “Verified Partner” affiliate link earns its place only after it has qualified on the certification.

How each certification standard is treated in filter matching
StandardScopeTreatmentWhat it means
NSF 53Health effectsQualifiesLead, VOCs, chromium, and (post-2022) PFAS. Satisfies a health hazard.
NSF 58Reverse osmosisQualifiesDissolved inorganics, arsenic, nitrate, and total PFAS. The only class that satisfies nitrate.
NSF 401Emerging compoundsQualifiesTrace pharmaceuticals and emerging contaminants. Satisfies a health hazard.
NSF 42AestheticAesthetic onlyTaste, odor, and chlorine — not a health claim. Isolated from hazard matching.
NSF 372Lead-free materialsMaterials onlyCertifies the device is made of lead-free materials — not that it removes lead.
P473Retired 2022RetiredFolded into NSF 53 and 58; never treated as a standalone qualification.

04Data & licensing

The sources behind every report

We build on public, auditable data and open standards. Here is exactly what powers the product and how each source is licensed.

  • EPA Safe Drinking Water Information System (SDWIS)

    Utility-reported contaminant occurrence, violations, and system inventory, retrieved via the Envirofacts REST API.

    US Government public domain
  • EPA community water system service-area boundaries

    The July 2024 release of explicit utility service-area polygons used for Tier 1 point-in-polygon resolution.

    US Government public domain
  • SimpleLab TEMM spatial layers

    Tiered Explicit, Match, Model layers that augment boundary coverage for Tier 2 proxy and Tier 3 modeled matches.

    MIT License (open source)
  • NSF/ANSI, WQA, and IAPMO certification registries

    Independent product certifications that back the filter database — curated and quarterly-audited against the source registries.

    Curated from public certification listings

Address-to-system mapping uses the July 2024 EPA community water system service-area boundaries, augmented by SimpleLab’s open-source TEMM layers. Certification data is curated from the NSF/ANSI, WQA, and IAPMO registries and audited quarterly. We report municipal engineering data only; WaterQualityLens is not a medical device.