Beauty Product Barcode Database: GTIN Data, Delivered

The query names a specific kind of machinery: a store where the printed code on the pack - the EAN/GTIN a camera snaps mid-aisle - resolves into a brand, a category and an ingredient list without anyone typing. In the personal-care slice of the Datadory catalog, exactly one dataset is built that way. Open Beauty Facts held 73,522 beauty products at the August 2026 verification pass, one row per product barcode, and it carries the top quality score in the slice: 10 out of 10 on the catalog rubric against a 7.81 mean across 1,744 cataloged datasets - one of just 145 records anywhere in the catalog to reach the maximum.

Everything else in the industry orbits the barcode rather than sitting behind it. Retailers key their shelves on their own identifiers, ingredient encyclopedias key on slugs, regulators key on substance names. That asymmetry is why this record works as the spine of any scan-driven pipeline: resolve the scan here first, then attach the joins. Datadory delivers the spine and the joins together, cut to the categories, brands and geographies you name.

The modeling extract verifies at 73,421 rows x 111 columns, and the fields worth knowing cluster into five jobs.

Identity. code, the EAN/GTIN itself, one value per row - the entire reason the dataset exists; product_name; brands plus the brand owner where recorded.

Taxonomy. categories_tags placing each product in a tree (en:hair-masks), countries_tags recording where it sells (en:france leads the observed values, which tells you the geographic story before any chart does), and packaging descriptors.

Formulation. ingredients_text, the complete INCI declaration - the single column that connects every scan to the regulatory and tolerability joins below.

Demand. unique_scans_n, the count of times that code was scanned in scanner apps - a popularity signal no storefront publishes and no survey panel can buy.

Quality. a 0-1 completeness score, states_tags marking editorial stage, and data_quality_errors_tags flagging known problems. Two conventions matter before you filter: products contributed without a printed code carry placeholders under the reserved 200 prefix (codes a scanner will never emit), and stub rows flagged ingredients-to-be-completed are common enough that the quality columns belong in every query rather than in downstream cleanup.

Three rows as they land in a delivery, flat and barcode-first:

# one row per scanned code, exactly as a delivery resolves it

code        : 0850018802796
brand       : Olaplex
ingredients : Water (Aqua/Eau), Cetearyl Alcohol, Dimethicone...

code        : 3600542613316
name        : Garnier Fructis Masque Nutrition Boucles Hair Food
categories  : en:hair-masks

name        : Sinful colors Gorgeous 804
brand       : Mirage Cosmetics
categories  : en:makeup

Read them honestly. First, the barcode is the primary key everywhere - the join key that survives renames, reformulations and re-shelving, which is why deliveries pin it rather than a display name. Second, the crowd's center of gravity is European drugstore and grocery, not the US prestige wall, so a French-language hair mask and a mass-market polish sit closer to the median than an American assortment would suggest. Third, quality varies row by row: a Kiehls entry in the same extract carries a completeness score of just 0.267, which is why the quality columns ship alongside everything else and why extracts arrive pre-filtered to the thresholds you set.

Scale: 73,522 beauty products worldwide at the August 2026 verification, mirrored as a typed 73,421-row x 111-column matrix for modeling work. The small difference between the two figures is two moments of the same production pass, not competing databases - we report both because someone will eventually notice.

Geography: contributions skew toward France and Europe, and the wider catalog publishes no country breakdown on its browse surfaces. Treat the corpus as globally broad and geographically lopsided at the same time.

Granularity: one row per product barcode at SKU level, with variants carried on the record rather than exploded into phantom rows.

A scan resolves into an ingredient string. Joined out, it becomes an analysis.

The joins themselves happen upstream of you: INCI spellings normalize onto one key before the merge, and every delivered table documents its grain.

• App builders turn a camera snap into brand, category and ingredient panel - the lookup layer underneath any scan-first beauty app, delivered as typed rows rather than engineered.
• E-commerce operators borrow the barcode-to-INCI structure to fix cold-start recommendations and enrich thin catalog entries with formulation context.
• Formulation and product-development teams read real declarations across thousands of live products and screen them against regulatory annexes before committing a reformulation direction.
• Competitive-intelligence teams track brand mix and category presence across a global corpus instead of resampling two national shelves.
• Data scientists train on a schema that holds between refreshes, with scan counts and review corpora as ready-made targets.
• Market researchers and consultants cite concrete assortment facts - SKU counts, category trees, country spread - when sizing beauty segments for clients.

Workflow detail lives on the persona pages: e-commerce operators, market researchers, data scientists and competitive intel product teams.

Start with a sample: name the categories, brands, price bands or geographies you care about and the extract arrives cut to them - real barcode-keyed rows in the shapes shown above, before any commitment.