Six Nines: What a 99.9999% Removal Report Proves, and What It Does Not

Search for bacteria removal on any e-commerce platform and 99.9% is the number you will see. Occasionally 99.99%. We hold a third-party laboratory report on one of our PES ultrafiltration cartridges recording a removal rate above 99.9999% - six nines. The gap between three nines and six is not marketing arithmetic: it is a factor of one thousand in what passes through the membrane. This article explains where that factor comes from, what the report actually measured, how to read any such document critically (including ours), and where the claim stops.

What the nines actually mean

Three nines, 99.9%, leaves one organism in a thousand. Four nines leaves one in ten thousand. Six nines leaves fewer than one in a million. Each nine is an order of magnitude, so the distance from 99.9% to 99.9999% is a thousand-fold reduction in breakthrough - not a rounding difference between two similar products.

That is also why the number cannot be evaluated on its own. A removal figure is only meaningful alongside the organism tested, the method used, and the physical form of the sample. Without those three, the percentage is decoration.

Three things to check before you believe a report

First, who ran it. A third-party microbiology laboratory carries weight that an in-house bench does not, because the buyer can ask the laboratory to confirm the report exists.

Second, what method. The report should name the standard and the test procedure, not just the outcome. A removal rate quoted without a method cannot be reproduced, and a claim that cannot be reproduced is not evidence.

Third, what was tested. A membrane module pulled off the spinning line and a finished cartridge that leaves the factory are different samples with different failure modes. A module result tells you the membrane works. A cartridge result tells you the membrane, the potting, the end caps and the seals all work together - which is the thing a buyer actually installs.

Our document answers all three. It was issued by Guangdong Detection Center of Microbiology (广微测, Gmicro Testing), report number 2026SP06971R01, sample received 20 July 2026 and completed 29 July 2026. The sample submitted was a PES ultrafiltration membrane filter cartridge - a finished cartridge, not a membrane module. The test item was the evaluation of disinfection efficacy for domestic drinking water.

The method is the Technical Standard for Disinfection (Ministry of Health, 2002 edition), Part 2, clause 2.1.4.1. Test water was prepared from pure water spiked with the challenge organism, Escherichia coli 8099, and the filtrate was sampled. Three independent runs, each with its own control.

Note the greater-than sign. In every run the test sample came in below 1 CFU per 100 mL - the limit of quantification - so the removal rate is a floor, not a measured point. The honest reading of this report is "at least 99.9999% under these conditions", not "exactly 99.9999%". A supplier who rounds that up to a precise figure has misread their own document.

Read the full four-page report 2026SP06971R01

NSF 42 is not a microbiological claim

A clarification that surprises many buyers: NSF/ANSI 42 covers aesthetic effects - chlorine, taste, odour and particulate reduction. It is not a microbiological removal certification and should never be presented as one. Bacterial removal is demonstrated by laboratory log-reduction testing, which is a separate exercise with a separate report.

This distinction runs both ways. A product holding 42 is not thereby proven to remove bacteria, and a product with an excellent removal report is not thereby certified for taste and odour. Both documents answer different questions.

Where hollow fibre earns the number

A hollow fibre membrane is a physical barrier. Water is pushed through the fibre wall under pressure while particles, colloids and bacteria larger than the pore are retained on the surface. Our elements are rated at 0.01 micron. Nothing is dosed into the water, so there are no disinfection by-products and no chemical residual to manage.

Outside-in and inside-out configurations are often ranked as if one were superior. They are not: the choice follows the feed water, the cleaning method and the installation. What decides whether a module performs is downstream of the membrane science - the potting, the glue lines, the end-cap seals. A single unbonded fibre or a glue void defeats a perfect membrane, which is why a cartridge-level test is worth more than a module-level one.

What our six nines do not prove

The result belongs to the cartridge that was submitted, tested under the conditions stated in the report. It is not a claim that every model in our range performs identically, and it is not a sterilisation claim: filtration removes organisms physically, but it does not replace disinfection where the source water is not potable, and it does not remove dissolved contaminants such as nitrate, hardness or metals.

Two further limits are printed on the report itself and worth repeating. It is a commissioned test on a single submitted sample, and the laboratory states that the results refer only to the sample tested - neither the material lot nor the production series is covered. And the challenge water was pure water dosed with the organism, which isolates membrane performance rather than reproducing a real feed: on turbid or organically loaded water, the interval between cleanings is what governs performance, not the pore size alone.

Nor does it remove viruses by assertion. A 0.01 micron barrier retains bacteria and protozoa; virus removal is a different test with a different result, and we will not quote one we do not hold.

Why publish the document at all

In B2B supply the expensive risk is not price, it is receiving something other than what was specified. A report a buyer can check - laboratory, method, sample, date - converts a claim into a traceable fact. That is the only reason to publish it, and the only test of whether it is worth publishing.

Ask us for the report against the model you are considering, and check the organism, the method and the sample form before you read the percentage. If the number survives those three questions, it is worth something. If a supplier will not send the document, the number was never the point.