gtin

GTIN is the join key between your catalog and Google's, Meta's, and Amazon's product knowledge graphs. With a valid GTIN, your listing is matched against the manufacturer's canonical product entry — same brand, same model, same specs, same global product page. Without one, your listing exists in isolation, competing in a less-confident matched set. The performance difference for branded products often runs 20-30% on click-through alone, because Google's confidence in showing your listing for "Sony WH-1000XM5" is calibrated by how well your listing matches Sony's canonical product.

GTIN format varies by region. UPC (12 digits, mostly US/Canada), EAN (13 digits, mostly EU/UK/Asia), JAN (8 or 13 digits, Japan-specific subset of EAN), ISBN (10 or 13 digits, books). All are GTIN variants; all are accepted everywhere. Don't convert format by region — channels handle the conversion internally. The exception is ISBN-10, which should be padded to ISBN-13 (prefix with "978") for channels that require ISBN-13 specifically.

The Excel-strips-leading-zeros problem deserves explicit warning. A valid GTIN like `0123456000018` opens in Excel as a number and becomes `123456000018` — instantly invalid. Any CSV passing through Excel between source and feed is at risk. The fix is to enforce text format on the GTIN column at every step, or skip Excel entirely in the pipeline. Many feed-generation systems use Excel as an intermediate format without realising this; you find out when the GTIN errors start appearing in Merchant Center.

Beyond structural validation, Google validates against GS1's database. A GTIN that passes the check-digit algorithm but isn't actually registered to any manufacturer gets flagged as "invalid". This catches placeholders that happened to pass structurally (`9999999999999` passes check-digit math but isn't registered to anyone). The two-layer validation is why `0000000000000` fails both — it's structurally invalid AND not registered.

Most retail products do have GTINs. The exceptions are genuinely identifier-exempt: custom-made jewellery, original art, vintage clothing, antiques, own-brand products where you're the manufacturer and never assigned one. For those, `identifier_exists: false` is the right answer. For everything else — yes, even when sourcing from a distributor who won't tell you the GTIN — the GTIN exists somewhere; you can usually find it on the manufacturer's website or in industry databases. Don't substitute a fake GTIN or a placeholder. Channels validate check digits, and bad data is worse than no data.

The other GTIN trap is variant-level uniqueness. A dress in five sizes and three colours has fifteen variants, and each has its own GTIN. Many feed pipelines mistakenly assign the parent product's GTIN to all variants, which causes "duplicate GTIN" warnings and confuses every downstream system. Each variant's actual barcode — the one printed on its specific packaging — is what should populate the gtin field for that variant.