Scan, understand, decide

The camera reads the barcode in your browser. The number is normalised to a single 14-digit form, so a UPC-A from an American pack and the EAN-13 from a European one find the same record. The check digit is validated, which catches a misread before it becomes a wrong answer.

That number is looked up in BeautySignal's own catalogue — one index probe, no third-party API in the path. If we don't have it, you are told plainly, it costs you nothing, and the barcode joins the list of what to add next.

The product's ingredient declaration — stored exactly as it was printed — is split into entries and each one matched against the ingredient reference. Anything we can't resolve is shown as unresolved rather than dropped, because a shortened list would be misleading.

Each matched ingredient brings its function, a plain explanation, and any regulatory position tied to the named instrument it comes from. Flags are then raised from those facts — not from a model, and not from a list of scary-sounding names.

No score. A number out of 100 would imply a judgement about safety the data cannot support. We report named dimensions and let you weigh them.

No verdicts. “Contains parabens — dangerous” is not a true statement. The true statement is longer and more specific, so that is what we show.

No AI in the answer. The analysis is a deterministic rules engine over structured data. The same product always produces the same answer, nothing is invented, and a scan costs a database query rather than an inference.

We say so. A product with no ingredient list says it has no ingredient list. An ingredient we don't hold says so. A gap in our data is never rendered as a clean bill of health. More about the database.