How we research water filters
By Daryna Klochko · Published · Updated Where the numbers come from
Certifier listings
Each filter is looked up in the public NSF and WQA listings by model name. “Certified” appears only with the standard and the substances in that listing, and the date we checked. A standard is not a certificate: any body with ANSI accreditation can certify to NSF/ANSI 53.
Brand pages and data sheets
Flow rate, cartridge life, cartridge price, warranty and return terms come from the brand's own pages and Performance Data Sheets. Every number on a model page names its source and the date it was checked. Reduction claims that aren't in a listing are shown as the brand's claims.
Other people's tests
Video reviews, lab reports and articles by other people. We record what was measured (flow, TDS, lab results), who measured it, with what and when, and show it beside the brand's figure. We don't test products ourselves, and a reading in someone's kitchen is not a certification.
Prices
From the store page linked next to each price, with the date it was checked. Prices and listings are rechecked once a month.
Owner reviews
From Amazon, other stores and the brand's own site. We read the ratings and the reviews, look for praise and complaints that come up again (leaks, flow dropping, cartridges running short), quote only word for word, and link to each review we quote. Reviews written for a free product or marked as paid aren't used, and neither are quotes that make health claims.
City and state water data
Violations and lead results from the EPA's SDWIS, PFAS results from UCMR 5, and each utility's annual Consumer Confidence Report, with the sampling date. The data describes a water system, not your tap.
Reading certifier listings
NSF/ANSI 42, 53, 58 and 401 are public standards. They don't certify anything by themselves: a certifier with ANSI accreditation checks a model against the standard and publishes a listing. The EPA names five such bodies: CSA Group, IAPMO R&T, NSF, UL and WQA. “Not listed by NSF” doesn't mean “not certified”, and “tested to NSF standards” on a product page doesn't mean certified at all.
A listing names the model and each substance separately. A filter listed for chlorine under NSF/ANSI 42 is not listed for lead, however the product page reads. That's why every reduction claim on our pages is tagged as certified or as a brand claim, and why we write “reduces”: a listing states a measured reduction, not a total. The contaminants or other substances removed or reduced by a water treatment device are not necessarily in your water.
Each filter gets one of four labels. The count is how many of the 137 filters in the database have it today.
| Label | What it means | Filters |
|---|---|---|
| Certified | A public listing by a certifier with ANSI accreditation (NSF, WQA, IAPMO R&T, UL, CSA) names this model, the standard and each substance. Only those substances count. | 92 |
| Certified, no health claim | The listing covers NSF/ANSI 42 or 177 (chlorine, taste, odor), or TDS alone under NSF/ANSI 58. TDS counts dissolved minerals; it says nothing about safety. Lead, PFAS or cysts on the product page are then brand claims, not certified claims. | 20 |
| Brand data only | The brand publishes its own lab report or data sheet, and no certifier listing covers the reduction claims. A listing for materials alone (NSF/ANSI 61 for components, 372 for lead content) isn't a filtration certification. | 25 |
| Listing not confirmed | The product page says “certified” or “tested to NSF standards”, and no NSF or WQA listing matched the model name when we checked. IAPMO, UL and CSA are the next catalogs to check. | 0 |
No page here says a filter is approved by the EPA or the FDA: neither agency approves home water filters.
How we use other people's tests
Flow, TDS and lab numbers on our pages come from video reviews, lab reports and articles that other people published. We go through each one, find the moment the reading is taken and record the number with its source. A model page then says how many independent tests were analyzed, and each number links back to the test it came from.
What a test needs before we use it
A test missing any of the five stays out of our tables, however good its numbers look.
Who ran it
Each test carries one of three labels, shown beside its numbers. All three kinds are used. The label is there so you can weigh them yourself.
What we record for each type
| Type | What we record | Usual method |
|---|---|---|
| Under-sink and countertop | Flow at the faucet in gpm; TDS before and after in ppm | Timed fill of a measured container; handheld TDS meter |
| Reverse osmosis | Gallons per day; drain water per gallon of filtered water; TDS before and after | Both outlets collected over a timed run; TDS meter |
| Pitchers and dispensers | Minutes to filter a full reservoir; TDS and chlorine before and after | Stopwatch; TDS meter; chlorine strips or a DPD kit |
| Shower filters | Free chlorine before and after; flow at the shower head in gpm | DPD kit or strips; timed fill of a bucket |
| Whole house | Pressure in psi and flow before and after the system; chlorine; hardness in gpg for softeners | Gauges on both sides of the system; chlorine strips; hardness test |
| Water tests | Days from mailing the sample to getting the report; results beside a second lab's report on the same water, if someone published one | Dates on the order and on the report |
For every type, lab results for named substances are recorded too, when the author sent water to a lab and names it.
TDS is not a safety measure
TDS (total dissolved solids) is the reading most videos show. It measures dissolved minerals and salts, which change how water tastes and looks. It says nothing about lead, PFAS or bacteria. So a TDS reading from a video is shown as a measure of minerals, and we never use it to call one filter better or worse than another.
When tests disagree
Readings from different homes differ because the water going in differs. Tap TDS, water pressure and the age of the cartridge all move the result. A measurement in someone's kitchen describes that water and that cartridge on that day. It isn't a promise for yours.
So tests are shown side by side, each with its source. Flow is the one figure we average: a model page gives the average of the readings, the highest one and the number of tests, and lists every test under them. TDS stays a pair for each test, before and after, and isn't averaged. Lab results are shown one report at a time, with the lab and the report date.
A reading far from the rest gets a second look at the video for a cause, such as a new cartridge that wasn't flushed or low house pressure. If the video shows one, the page says so beside the number. If it doesn't, the reading stays in the list.
For lead, PFAS and other substances that affect health, the claim on our pages comes from the certifier listing. Someone else's lab result for the same substance is extra information, never a stand-in for a listing. When a test and a listing disagree, the page shows both.
How each number is credited
Each number names who measured it and links to the source: a video at the minute and second of the reading, a report by its number and date, and the lab when the author names one. The same sentence carries the line “This is not a certification.” Where nobody has published a reading, the page says “No independent test found as of” and the date, rather than leaving the cell empty.
<channel> measured <N> gpm at the kitchen faucet (video, <mm:ss>). This is not a certification.
We describe tests in our own words, quote a line at most and link to the original. Videos aren't embedded. A screenshot from a video shows the reading itself, has no faces in it and keeps the channel's watermark as published.
How the score works
Every model is scored on five criteria. Each criterion gives 2, 1.75 or 1.5 points, so the score out of 10 runs from 7.5 to 10. There are no weights. Each criterion is worked out from the fields in our database, and the category median is taken over models of the same type. The word beside a score follows fixed steps: 9.5 and up is Outstanding, 9 Excellent, 8.5 Very good, 8 Good, and anything lower Decent.
Owner reviews: “mostly positive” means a store other than the brand's own site shows at least 25 ratings averaging 4.3 or higher, and no such store averages under 4.0. “Repeated complaints” means one of two things. On such a store, at least 10 reviews about leaks, flow or breakdowns are negative, and they make up 40% or more of the reviews on that topic and 2% or more of all its ratings. Or at least 3 of the reviews we read, and 5% or more of them, describe a unit that stopped working. Ratings on the brand's own site alone give 1.75 points, because the brand decides which reviews it shows. So do fewer than 25 ratings on other stores, and so does a model whose reviews we haven't gathered yet.
Independent flow and TDS readings are not part of the score. They are shown beside the brand figure.
Water filters
| Criterion | 2 points | 1.75 points | 1.5 points |
|---|---|---|---|
| Certification | Listed by a certifier with ANSI accreditation for a substance that affects health (NSF/ANSI 53, 58 or 401, or NSF P473), with the listing link and the date we checked. TDS alone doesn’t count | Listed for chlorine, taste or odor only (NSF/ANSI 42 or 177), or for TDS only (NSF/ANSI 58) | No listing: a brand lab report or a listing for materials only (NSF/ANSI 61 or 372) is not one, nor is “certified” on a product page with no listing we could find |
| Published specs | Cartridge life and flow rate both published as numbers. For reverse osmosis, flow means gallons per day and the drain ratio. For pitchers, flow means minutes to filter a full reservoir | One of the two | Neither |
| Cost of ownership | Cartridge cost per year at or below the category median | Up to 1.5× the category median | More than that, or no replacement interval published |
| Warranty and returns | Warranty at least as long as the category median, and returns of 30 days or more | One of the two | Neither, or final sale |
| Owner reviews | Mostly positive | Mixed, or not gathered yet | Repeated complaints about leaks, breakdowns or dropping flow |
Water tests
| Criterion | 2 points | 1.75 points | 1.5 points |
|---|---|---|---|
| Lab accreditation | — | Samples analyzed by an accredited lab | Home strips or a meter, no lab |
| Published specs | Number of substances analyzed and turnaround both stated | One of the two | Neither |
| Price | At or below the category median | Up to 1.5× the category median | More than that |
| Warranty and returns | Same as filters | Same as filters | Same as filters |
| Owner reviews | Same as filters | Same as filters | Same as filters |
What affects rankings
The score comes only from the five criteria above. Whether a brand pays us a commission plays no part in it, and brands can't pay for placement.
The commission does affect the order. By default, lists show models with a “Check price” button first and the rest after them, and our picks come only from models with a button. Inside each group, models go by score, and when two tie, the cheaper one comes first. To see a whole list by score alone, sort it by “Highest score”.
Cost of ownership
Cost of ownership on our pages is what the replacement cartridges cost, worked out from the brand's own figures in two ways: per year and per gallon. The purchase price isn't part of it and is shown on its own.
The sum covers every part the brand says to replace: pre-filters, the main cartridge, the reverse osmosis membrane and post-filters. Cartridge prices are the brand's own store prices on the date shown with them. Many brands rate a cartridge in both months and gallons and say to replace it at whichever comes first. The yearly figure uses the months, so a household that runs through the gallons sooner pays more than that.
When a brand publishes gallons but no replacement interval, the page gives the cost per gallon, says the yearly cost isn't published, and the cost criterion gets 1.5 points. Drain water from reverse osmosis isn't priced in, because it depends on your water rates. The drain ratio is shown beside the cost instead. Installation, and electricity for systems with a pump, aren't included either.
Evidence levels
When a page says what a substance in water does or what a filter type can do about it, it links the source and says what kind of source it is. Across a whole topic the evidence gets one of four levels.
If a claim has no public source behind it, the page says whose claim it is, such as “the brand says” or “owners on Reddit report”.
Updates
Prices and certifier listings are rechecked once a month, on the 1st. New independent tests are added as we find them. A page's “Updated” date moves only when something on it changes: a price, a spec, a new test or a new model. Every price shows the date it was checked, and every listing the date we looked it up.
How we make money
Some links on this site are affiliate links: we may earn a commission when you buy through one. The store or its affiliate network credits that sale with its own cookie, as the privacy policy explains. The commission doesn't change a score, and brands can't pay for placement. What it does change is the order of a list, as What affects rankings explains.
Affiliate disclosureCorrections
If a number is wrong or out of date, tell us through the contact page. We check it against the source, fix the page and move its “Updated” date.
Report a numberHow to cite
You're welcome to quote figures and short passages from WaterJury with a link to the page you took them from. A format that works:
WaterJury, “<page title>”, <page URL>, accessed <date>.
Most figures on our pages come from somewhere else first: a brand, a certifier, the EPA or an independent tester. If a number matters to your piece, cite that original source as well. Each one is linked beside the number. Copying whole tables or the database needs permission, as the terms of use explain.