Where the information comes from

How a source is chosen

Sources are ranked by how close they are to the original evidence. Regulatory registers, standards bodies and peer-reviewed literature come first. A manufacturer is an acceptable source for facts about its own product and for nothing else. Content aggregators, SEO-built listicles and sites that restate figures without attributing them are not used at all — a number with no origin cannot be checked, and repeating it would only launder it.

Every citation records the specific claim it supports and the date we retrieved it, so a stale fact can be found later rather than quietly persisting.

Datasets we use

CosIng — the European Commission’s cosmetic ingredient database. Open data. It provides INCI names, function classes and the EU regulatory restrictions on an ingredient. It arrives with the ingredient checker and will be the backbone of the ingredient pages.

PubChem and FDA registers — chemical identifiers, for matching an ingredient across the different names it appears under.

The plain-English explanation of what an ingredient is there to do is written by us, reviewed by a person, and is not taken from any of these. Where a rating scale exists elsewhere, we do not reproduce it: those scales belong to whoever built them, and we would be passing on a judgement we had not made.

What we do not have

Live retail prices, second-hand values and review corpora are all unavailable to us on terms we can legitimately use. The tools that would depend on them are therefore not built. When a workable source exists, the tool follows; the order is never the other way around.

Import duty rates are a partial case: there is no free, reliable global tariff feed, so the import calculator will ship with a small set of lanes whose rates have been verified against the relevant customs authority, and will say plainly when your lane is not one of them.

The rules behind all of this are on the methodology page.