GTIN and Product Attribute Cleanup for Feed Disapprovals

What the number actually is: Across the last four catalogues we took on, between 6 and 31 percent of items carried a GTIN that passed its check digit and still belonged to somebody else's product. Validation tools miss this, because a well-formed identifier is not the same as an owned one: the check digit is arithmetic, ownership is a company prefix registered to a named party. When a supplier recycles a code from a discontinued line, your item inherits another brand's price history and competitive set inside Shopping, and the disapproval that eventually lands is the least expensive part of the damage. How we work through a catalogue: We resolve every identifier against ownership rather than format. Prefixes are checked against the registering party, so a code that is arithmetically valid but registered to an unrelated company is flagged instead of passed. Multipacks and bundles are separated out early, because a three-pack you assemble is a new trade item that needs its own identifier plus the multipack or bundle declaration. Borrowing the single-unit code for it is the most common self-inflicted mismatch we see in grocery, pet and household. Genuinely unidentified goods, meaning your own manufacture, custom or made-to-order lines, get the brand and MPN pair with the identifier-exists declaration set to false. That declaration is a statement of fact about the product, not a switch for silencing warnings, and an account that has been using it that way is a remediation job before it is a feed job. Variant structure is the other half of the work. Items grouped under one item group need consistent colour, size, material and pattern values drawn from a controlled list rather than free text typed by whoever loaded the product that afternoon. Where the market requires it, unit pricing measure and base measure are derived from pack data instead of parsed out of the title, which is where most automated attempts quietly produce nonsense. Where the fix goes: Into the source. Rules written into a feed tool are a tourniquet. They hold until somebody adds a category, and they do not travel to Amazon, eBay or your retailer portals, each of which enforces identifiers under its own rules and its own exemption process. We write the field mapping and the validation rules against your PIM or ERP, and we leave behind an export check that fails the build when an item goes out without an owned identifier. Your data team runs it without us. What we will not do: We will not generate identifiers. Codes come from GS1 under your own company prefix or from the brand owner, and any consultancy offering to supply them is selling you a future recall of your listings. We will not set identifier-exists to false across a catalogue to clear disapprovals. We will not take this on as a one-off scrub when the buyer has already decided the PIM is untouchable, because the same items fail again within two quarters and we will have been paid for the interval. Who should not buy this: Catalogues under roughly two thousand active items, with one merchandiser and a stable supplier list. That is a fortnight of careful work by somebody who already knows the range, and paying us is worse value than paying them overtime. Resellers who cannot get pack data out of their suppliers should fix the supplier onboarding form first. We are happy to review that form and stop there.