How the grade works
Scoring method v2.0. Last written 16 November 2026, from the WaterScore methodology used in the Florida pilot.
WaterScore gives one overall grade built from three parts. Each part also has its own score, grade, confidence, and reasons.
| Part | Weight | What it measures |
|---|---|---|
| Tap-water test results | 50% | Measured contaminant levels (lead, copper, nitrate, arsenic, PFAS, disinfection byproducts TTHM/HAA5, radionuclides, other chemicals) compared with legal limits |
| Compliance record | 25% | The utility's Safe Drinking Water Act violation history (SDWIS, with ECHO flags as a gap filler) |
| Environment | 25% | Nearby Superfund sites, impaired waters, fish advisories, red tide, and other layers |
None of this is a lab test of the tap at a given house, and none of it is medical advice. The test results are samples the utility or state took at the plant, at entry points, and in the distribution system.
Missing or unavailable data lowers confidence. It is never treated as a clean record. A layer or file that could not be loaded is not scored as “nothing found.”
Overall grade
Overall = 0.50 × results + 0.25 × compliance + 0.25 × environment, rounded.
A 90–100, B 80–89, C 70–79, D 60–69, F 0–59.
If a part cannot be scored (no measured results on file, no violation extract, or every environmental layer unavailable), it is left out and the weights of the other parts are scaled up so they still sum to 100%. The panel marks the part “excluded” and shows the weights actually used. Leaving a part out also lowers overall confidence one step.
Overall confidence is the lowest confidence among the parts that were scored, minus one step when a part had to be left out.
Tap-water test results (0–100)
Starts at 100. Each measured contaminant is turned into a ratio: measured level divided by its limit (EPA maximum contaminant level or lead/copper action level; for organics, the limit Florida DEP lists for the sample). Units are converted (mg/L, µg/L, ng/L); pCi/L is compared as is.
| Measured level as a share of the limit | Points taken off (before weighting) |
|---|---|
| Under 20% | 0 |
| 20% to 50% | 0 up to 5 |
| 50% to 100% | 5 up to 20 |
| At or over the limit, enforceable now | 40, plus 10 per extra multiple of the limit, capped at 55 |
| Over a limit that is not yet enforceable (PFAS) | 22, plus 10 per extra multiple, capped at 32 |
The worst contaminant counts in full, the second at 60%, the rest at 40%, so many small readings cannot add up to a failing grade by themselves, and one bad reading cannot hide behind clean ones.
What level is used for each contaminant
- Lead and copper: the most recent 90th-percentile result from EPA SDWIS (every 1 to 3 years). Action levels: lead 15 µg/L, copper 1.3 mg/L. EPA's 2024 rule lowers the lead action level to 10 µg/L from November 2027. Lead has a health goal of zero, so a 90th-percentile result of a few µg/L still counts for something.
- TTHM and HAA5 (disinfection byproducts): the highest location average over the 12 months up to the newest sample, which mirrors how the legal limit (80 and 60 µg/L) is applied. If Florida DEP has no result, UCMR 4 HAA5 is used as a fallback.
- PFAS (UCMR 5, 2023–2025): the average of the entry-point samples, with non-detects counted as zero. PFOA and PFOS are compared with 4.0 ng/L. EPA's 2024 rule makes that enforceable in April 2029; in May 2026 EPA proposed a two-year extension to 2031, so it is scored on the lower “not yet enforceable” scale. PFHxS, PFNA, and HFPO-DA (10 ng/L) are scored the same way and are marked as proposed for rescission. Other PFAS and lithium have no individual federal limit and are listed, not scored.
- Everything else from Florida DEP (inorganics, radionuclides, volatile and synthetic organics): the highest detected value in the sample years on file.
- Not detected: listed by name with how many samples were taken. A “not detected” lab report means below the lab's reporting level, not zero.
Secondary (taste and odor) standards, turbidity, hardness, pH, sodium, and similar are not scored.
Confidence. High needs recent results (within 6 years) in all four key groups: lead sampling, disinfection byproducts, inorganic chemicals, and PFAS. Three groups is medium, two or fewer is low. The reasons name which groups are missing. If there is no results data for the system at all, the part is “not available” and reweighting applies.
Compliance score (0–100)
This part is about the drinking-water utility's Safe Drinking Water Act record. It starts at 100 and only uses the SDWIS violation extract for the matched community water system, plus EPA ECHO flags when that extract has no open row but ECHO says a problem is still open. Private wells are not scored.
Health-based violations count more than monitoring or reporting violations. Open (unaddressed or addressed, but not resolved) violations are not reduced for age, because they are still open.
| Kind | Unaddressed | Addressed (still open) | Cap |
|---|---|---|---|
| Health-based | −22 | −14 | −66 |
| Monitoring or reporting | −5 | −3 | −20 |
| Other | −3 | −2 | −9 |
Resolved violations use recency decay. The date is the return-to-compliance date when it is present, otherwise the end date, otherwise the compliance-period start. The full resolved penalty applies at age zero and falls in a straight line to zero at the end of the window. A health-based violation resolved eight years ago does not lower the score. A monitoring violation resolved five years ago does not lower the score. If the date is missing, the full recent resolved penalty is used, so a missing date is not treated as “old enough to ignore.”
| Resolved kind | Full recent penalty | Window | Cap |
|---|---|---|---|
| Health-based | −8 | 8 years | −16 |
| Monitoring or reporting | −2 | 5 years | −6 |
Archived violations are ignored. They no longer count in EPA’s compliance status. Older resolved violations outside the window are also left out of the local file.
If the extract has no open violation, but ECHO flags a health-based violation, open violation points, or a current violation, that flag is used so a missing row does not look like a clean record (−22, −12, or −8).
If the violation extract itself is missing, there is no compliance score (not 100) and confidence is low.
Confidence. High means the loaded violation extract was used without an ECHO gap fill. Medium means an ECHO flag had to stand in for a missing row, or the as-of date was missing. Low means there is no compliance score (no extract, no matched utility, failed geocode, or outside the Florida pilot).
Environmental rating (0–100)
This rating is separate from the other two parts. It starts at 100 and only uses layers that actually returned a result. A successful empty search (“none”) can be used. An unavailable or missing layer cannot. If every scored layer is unavailable, the rating is omitted, not set to 100.
Distance matters. A site next door counts more than one near the edge of the search radius. If distance is missing, the full nearby penalty is used, so a missing distance is not treated as “far enough to ignore.” Extra sites after the worst one add a quarter of their own deduction, with a cap.
- Superfund / SEMS (10 miles). An NPL (or proposed NPL) site starts at −30 within 1 mile and falls to 0 at 10 miles. A SEMS facility not listed as NPL starts at −10 on the same curve. Deleted NPL sites use the smaller SEMS scale. Cap: −40.
- Impaired waters, ATTAINS (5 miles). IR category 5 starts at −16 within 0.25 miles. The distance weight is squared, so the deduction shrinks faster than a straight line: a water about a mile away counts much less than one next to the property, and the deduction falls to 0 at 5 miles. IR category 4 starts at −12. Other impaired waters start at −8. Extra waters after the worst one add a quarter of their own deduction. Cap: −16. Impairment is a waterbody assessment, not a tap test.
- Category caps. No single environmental category can take the rating below 60 on its own, so one layer cannot by itself produce a grade F. The largest cap is −40 (Superfund). Impaired waters are capped at −16. A failing environmental rating (F, 0–59) can only come from more than one scored category.
- Healthy Beaches (15 miles). The public layer lists sampling sites and Florida Department of Health links, not the current advisory or bacteria reading. Nearby sites do not change the rating. Confidence notes that the advisory itself was not in the data.
- Fish consumption advisories (15 miles). An advisory starts at −8 within 1 mile and falls to 0 at 15 miles. Cap: −12. This is meal advice, not a tap test. The Florida mapper layer was last edited in October 2018, which is called out in confidence.
- Red tide, Karenia brevis (15 miles, about the last 8 days). High abundance starts at −22 within 2 miles and falls to 0 at 15 miles. Medium is −14, low is −6, very low is −2. “Not present” or background does not lower the rating. No samples in range also does not lower the rating, but confidence notes that a missing sample is not a confirmed clear result.
- Groundwater (1 mile). Water Quality Portal wells are locations, not sample results for this address. They do not change the rating. An unavailable layer lowers confidence and is not treated as clean well water.
Confidence. High means every layer used in the rating returned a result. Medium means one or two layers were unavailable, or two or more data-quality caveats apply (for example, beaches without advisory text and a 2018 fish-advisory layer). Low means three or more layers were unavailable, or no rating could be produced.
The lookup also returns the top contributing reasons, in plain language, with the point change for each one.
What this version does not do
- It does not test the water at the house. Test results come from the utility's plant, entry points, and distribution system, not from your tap or your building's plumbing.
- It does not read utilities' annual Consumer Confidence Reports (PDFs). They are a good future source of detected-contaminant ranges.
- It does not apply health guidelines stricter than legal limits, except through the zero-goal weighting of lead and the PFAS limits, which were set close to health goals.
- Florida DEP's chemical files are published by sample year. Files for 2020, 2021, 2022 (partial), and 2024 are used; 2023 and 2025 are not published. Many contaminants are only sampled every 3 to 9 years, so a result can be several years old. The panel shows the year.
- It does not use lead service line inventories, private wells, live beach bacteria results, or real-time sensors. See data sources for the full list.
- Service-area boundaries can be incomplete or modeled, so a house on the edge of an area can be assigned to the wrong utility or to no utility.
Questions about the method: waterscore.app@gmail.com.