How we verify
Every specification published here is traced to a document you can open. The source is cited on the page beside the figure it supports, and the page carries the date the figure was last checked against that document. Where no source can be found, no figure is published.
The rule
One figure, one source, one date. A number that cannot be pointed at a published document does not go on a page, however plausible it looks and however many other sites repeat it. This is the whole method, and everything below is the machinery for keeping to it.
What counts as a source
Sources are ranked, and a page uses the highest one available for each figure:
- Standards. ISO, ANSI and their national equivalents, cited by number and edition year — ISO 9626:2016, not "ISO". An edition matters: values move between them.
- Regulators and government databases. The FDA's device databases, the NIH's AccessGUDID registry, the Code of Federal Regulations, OSHA, USPS Publication 52, and the environmental and health agencies of individual states.
- Manufacturer technical documentation. Data sheets, instructions for use, and catalogue specifications published by the company that makes the product — not its marketing copy, and not a distributor's product listing.
- Retail listings, only for prices and availability, and only where the page says outright that a price is a retail figure observed on a date. Marketplace prices move hourly and are labelled as such.
What is not a source: another reference site, a content farm, an AI assistant's answer, a forum post, or a figure repeated widely enough that it looks settled. If the only support for a number is that everyone says it, it does not appear here.
How a page is made
Sourcing comes first and it is done by hand. Before a word of a page is written, the sources for it are collected into a packet — each figure with the document it came from and the URL that document lives at. A page that cannot be sourced inside a fixed time budget is dropped rather than written thinly, and several targets have been dropped that way.
The page is then drafted from that packet and from nothing else, and checked back against it by several independent passes that do not share the drafter's assumptions. Any statement that cannot be tied to a line in the packet is a defect: pages are held below a hard ceiling of unsupported claims, and a page over that ceiling does not ship. Several have been rewritten three and four times to get under it, and one was restructured entirely because its subject could not be reported honestly in the shape it was first given.
Mechanical checks run over every page before it can be published: that it invents no internal link to a page that does not exist, that it carries no placeholder domain, that no affiliate link can escape the disclosure, that its structured data is well formed, and that it makes no claim of continuity with the former business that once used this domain. A page failing any of them does not build.
Two further checks are run by hand rather than by the pipeline, and it is worth being exact about which. Where a page's table is a published standard, that standard is transcribed once into a dated reference file and the file is checked against the document; every page drawing on it is then verified against the file cell by cell, which is a script, run deliberately, on the pages that need it. Source URLs are confirmed by opening them on the verification date, and re-checked in bulk between batches — also by hand. Neither is automatic, and a page is not blocked from building by either.
Drafting and checking are automated; publishing is not. No page reaches this site without a person reading it against its sources and deciding to publish it.
When sources disagree
They disagree more often than you would expect, and the disagreement is usually the most useful thing on the page. The rule is to report both figures, name both publishers, and say which one governs and why — never to average them, quietly prefer one, or pick the tidier number.
Two worked examples, both live on this site. Nipro lists the outer diameter of a 29-gauge needle as 0.34 mm, while ISO 9626:2016 places 29G at a 0.33 mm metric size with a permitted band of 0.324–0.351 mm; the manufacturer's figure sits inside the standard's band, so both are correct and the standard is cited for dimensional reference. Component Supply publishes hub colours that contradict ISO 6009 outright — 25G is red in one and orange in the other — because the two systems cover different products, which is stated on the page rather than resolved in favour of whichever came first.
What the verified date means
It means a person opened the cited documents on that date and confirmed the figures on the page still matched them. It is not the date the page was published, edited, or reformatted. The date is moved only when a value or a source has actually been re-checked — a date that advances every time a page is touched would tell you nothing, so ours does not.
What we do not publish
- Medical advice of any kind. No dosing, no titration, no injection technique, no guidance on whether a product suits you. Those belong to a licensed prescriber or pharmacist. This boundary is enforced on every page, not applied case by case.
- Part numbers from the former store's catalogue. Nobody can order them from anyone, and printing them on this domain would imply a continuity that does not exist. Archived capacities and fitments are published where they are genuinely useful; the SKUs are not.
- Figures we could not check. A gap on a page means the publisher does not state it. Empty is honest; a plausible guess is not.
Known limitations
Two, stated plainly because a method page that lists none is not describing a real method.
Sources drift. Manufacturers reorganise their sites and agencies rename pages, so a cited URL that resolved on the verified date may later redirect or break. Three source URLs across the corpus are already permanent redirects rather than direct hits. Nothing here checks them continuously — the re-check happens between batches, by hand — so a link on this site is warranted to have resolved on the date beside it, and not warranted to resolve today. A source that has genuinely disappeared is replaced, or the figure comes down.
Coverage is uneven by design. This site publishes what can be sourced well, not what is most searched for, so there are obvious topics it does not cover and will not cover until the documentation exists to do it properly.
Corrections
If a figure is wrong, or a source has moved or been withdrawn, it gets fixed against the source and the page's verified date is updated to the day the correction was checked. A correction that changes a published value is made on the page itself rather than silently.
Page last verified: