Every water filter we have assessed so far arrived with a percentage on the box. Almost none of them arrived with the three things that make a percentage mean anything: which contaminant, at what starting concentration, and for how many litres.
This is not usually dishonesty. It is that the useful version of the claim is boring and the box is small. But it does mean the number on the front is close to useless for choosing between two products, and the information you need is somewhere else.
The four questions
These are the questions we put to every filtration brand before a product can be listed. You can ask them yourself, and the answers are usually in the datasheet rather than on the product page.
1. Certified by whom, or just tested?
There is a real difference between certified to a standard and tested to it. Certification means an accredited third party — NSF, the Water Quality Association, IAPMO — ran the protocol, audited the factory, and continues to check. “Tested to NSF/ANSI 53” can mean the brand paid a lab once, for one sample, and the result lives in a drawer.
Both are better than nothing. Only one of them is checkable: certified products appear in the certifier's public listing database, searchable by brand and model. If a product claims certification and is not in the listing, that discrepancy is the story.
2. Certified for what?
Certification is granted per contaminant, not as a blanket. A filter can be certified under a standard for exactly one substance and still, truthfully, say it is “NSF/ANSI 53 certified”. The standards split roughly like this:
- NSF/ANSI 42 — aesthetic effects. Chlorine taste and odour, particulates. This is the easy one, and the one most cheap filters hold.
- NSF/ANSI 53 — health effects. Lead, cysts, VOCs and similar. Each claim is separately certified.
- NSF/ANSI 58 — reverse osmosis systems specifically.
- NSF/ANSI 401 — incidental contaminants: pharmaceutical residues, pesticides, and other compounds that turn up in trace amounts.
A jug certified under 42 for chlorine and nothing else is a perfectly good jug. It is not a lead filter, and a percentage on the box will not tell you which one you have bought.
3. Certified for how long?
Every certified reduction claim comes with a capacity, stated in litres or gallons. The filter meets its claim up to that volume and is not tested past it. A cartridge rated at 400 litres in a household getting through 6 litres a day is a two-month cartridge, whatever the packaging says about “up to six months”.
A filter past its rated capacity is not a weaker filter. It is an untested one.
This is the most common practical failure we see, and it is not a product defect. It is a mismatch between a capacity in litres and a replacement schedule in months, sold to households of very different sizes.
4. Tested at what flow rate?
Contact time is doing much of the work in a carbon filter. The certified figure is measured at a specified flow rate; run the same cartridge faster and performance drops without any visible change. This is why an under-sink system and a jug using nominally similar media can produce genuinely different results, and why “fast flow” as a selling point deserves a second look rather than a first one.
What this means for a UK kitchen
Mains water in the UK is regulated and routinely tested, so the honest starting point is that most people here are not filtering to make water safe. They are filtering for taste, for chlorine, for hardness, or because of the plumbing between the main and the tap rather than the main itself.
That last one matters. Lead is a pipe problem, not a supply problem, and it is specific to the property. If lead is your reason for filtering, the useful step is finding out what your supply pipe is made of before choosing a cartridge — and then buying a filter with a lead reduction claim certified under NSF/ANSI 53, not a general purpose one.
How we are applying this
Every filtration product in the shop is listed with the standards it actually holds, per contaminant, and its rated capacity in litres. Where a brand could not produce a certification listing for a claim on its own packaging, the product is marked Did not pass and the reason is published with it.
Two systems are currently marked in testing. In both cases the paperwork is in order; what we are still measuring is how the claim holds up nearer the end of the rated capacity, which is the part nobody tests in a review.
A limit worth stating. We are not an accredited laboratory. Where we describe our own measurements, we publish the method, the number of samples and what we could not control. Where we describe a certification, somebody else did the work and we checked the listing.