Our Methodology

How we source data, what each number means, and where the data runs out. Figures below reflect the 2026 Q2 EPA snapshot.

Where the data comes from

Source	What we take from it
EPA SDWIS via EPA ECHO	Violations and enforcement, lead and copper 90th-percentile results, system characteristics, and state sanitary survey findings.
EPA UCMR5	PFAS sampling, 2023–2024. The fifth Unregulated Contaminant Monitoring Rule cycle, covering systems serving 3,300 people or more.
EPA UCMR3	Earlier PFAS sampling, 2013–2015, retained for trend work only. Its reporting limits were 10–90 ppt against UCMR5's 3–5, so a UCMR3 non-detect means "below about 20 ppt", not "absent".
US Census	Place and township boundaries, and the block-level assignments used to work out which systems serve which community.

We do not use Consumer Confidence Reports as a data source. CCRs are self-reported by utilities and not independently verified. SDWIS is federally mandated and audited.

MCL vs. MCLG — the distinction that matters

MCL (Maximum Contaminant Level) is the legally enforceable limit. Exceeding it puts a utility in violation of federal law and obliges it to notify customers and take corrective action.

MCLG (Maximum Contaminant Level Goal) is an aspirational target, set at zero for many contaminants including lead. Because it is zero, virtually every water system in America can be made to look as though it has a "lead problem" when measured against it.

What the difference looks like in practice The MCLG for lead is zero, so any detectable lead is above it. Of the 89,152 water systems with a lead result on file, 65.8% show some detectable lead and would appear to have a "lead problem" measured against the goal. Measured against the 15 ppb action level — the number that carries legal force and triggers mandatory corrective action — 5.44% do. Grading against the goal makes twelve times as many systems look unsafe as grading against the standard.

We grade against enforceable standards and action levels only. Where a contaminant has no enforceable limit, we say what was measured and leave the judgment to you.

Lead

SDWIS reports lead as a 90th-percentile result in mg/L; we convert to parts per billion. The federal action level is 15 ppb — a treatment trigger, not a legal maximum, and not an MCL. It is measured across sampled taps, not at any one home.

A reported zero means "not detected", not "no lead" A 90th-percentile result of 0 means the sample fell below the laboratory detection limit. We render it as Not detected and never as "0 ppb", which would tell you something nobody measured. Nationally 33.8% of communities with a lead result report one, but the rate is a state filing convention rather than a water quality signal — it runs from 0% in Tennessee to 63% in California. For that reason we exclude non-detects from any median and publish the count separately.

We show the most recent result per system. Where none exists we say so: 1.2% of community pages have no lead result on file.

PFAS

EPA set the first federal PFAS limits in 2024: 4 parts per trillion for PFOA and PFOS, and 10 ppt for PFHxS, PFNA and GenX. Comparison uses EPA's own significant-figure rounding, so a 2-significant-figure limit compares at 0.00405 µg/L.

The limit applies to a running annual average at each sampling point Not to any single sample. A system can exceed at some sampling points and meet the limit at others, and a one-off high reading is not a violation. We report "exceeds at N of M sampling points" rather than a single system-wide figure, because averaging across a system hides where the problem is and quoting a peak against an average-based limit overstates it.

UCMR5 required monitoring only from systems serving 3,300 people or more, so coverage is uneven by design: 45.3% of communities have PFAS results. Where a system was never sampled we say not tested and explain why. That is not evidence of absence, and we do not treat it as such. Of communities with results, 2,050 have PFAS above a federal limit.

State inspections

States conduct sanitary surveys of every community water system, grading eleven parts of the system separately — treatment, distribution, storage, source water, management and others. 100% of the communities we cover have an inspection on record since 2020, making this the best-covered dataset on the site.

A category marked not evaluated or not applicable was not assessed on that visit. It is not a pass, and we never render it as one.

How grades are calculated

Each system starts at 100 points. Deductions:

Finding	Deduction
Any health-based violation currently open	25
Health-based violations since 2015 — 3 or more / 1–2	14 / 6
Lead at or above the 15 ppb action level	25
Lead between 5 and 15 ppb	8
Copper at or above the 1,300 ppb action level	10
PFAS above a federal limit — at 2× or more / below 2×	30 / 20
Significant inspection deficiencies — 3 or more / 1–2	18 / 10

A is 90 or above, B 78, C 62, D 45, F below that.

A system we cannot verify cannot earn an A If PFAS or lead has never been measured, the grade is capped one below the top. Absence of testing never helps a system's grade — otherwise a system nobody examined would outrank one that was tested and found clean. It is not penalised either, because sampling thresholds are outside a utility's control. Grades therefore state what they were based on, and 28,893 of our graded communities carry that cap.

Why we do not rank states against each other

Lead reporting completeness ranges from under 20% of systems in some states to 100% in others, and PFAS coverage is set by a population threshold rather than by water quality. A state that tests more finds more.

Ranking states on either measure would sort them by reporting practice and system size, not by how clean the water is. We rank only within a state and within a comparable system-size band, and we publish coverage percentages alongside any median so the basis is visible.

Population figures

We derive population from census blocks inside a system's mapped service area, not from the figure utilities self-report. Self-reported counts double: when a wholesaler sells water to another system, both report the same residents, and active community water systems together self-report more people than live in the United States.

The trade-off is that boundary-derived figures run about 12% below self-reported ones, and further where a system's water reaches beyond its mapped boundary. We accept an undercount with a floor over an overcount without one.

Service areas are approximate Boundaries are drawn from EPA and state service-area data and mapped to census blocks. They are close, not exact, and they can lag changes on the ground. If you need to know with certainty which utility serves a specific address, contact the utility — a map cannot answer that question reliably and we do not pretend otherwise.

Where a community is only part of a larger service area, we show both figures: how many residents here are served, and how many the utility serves in total. The grade and the peer group are based on the utility, because that is the entity being measured.

What we cover

We grade 49,378 active community water systems — those serving the same residents year-round — across 30,524 communities and 3,129 counties. Violations are shown from 1 January 2015 onward, covering 1,453,783 records; older entries are retained for trend work but are too stale for a current assessment.

We exclude transient systems such as restaurants, filling stations and campgrounds, which serve passing traffic rather than residents.

Limitations

SDWIS is only as current as utilities' federal reporting; small systems in particular can lag.
PFAS was never sampled at systems serving fewer than 3,300 people, which is most systems though a small share of the population.
SDWIS carries violations plus lead and copper. It does not carry routine detection results for most contaminants, so we cannot tell you everything that is in your water — only what has been measured and what has been enforced.
Water hardness is not in SDWIS and we do not publish it.
Boundary-derived population can understate systems whose water travels beyond their mapped area.
Federal standards only. Several states set stricter limits that are not reflected here.
A community served by several systems is reported on the one serving most of its residents; the page names which.

Questions about our methodology?

Get in touch. Our figures are derived from EPA records, so in most cases we can trace a question back to the underlying data — and we'd rather hear about something that looks off than not.