Catalogue and feed repair for Meta, TikTok and Pinterest

Most catalogue problems that cost money never appear as errors. Commerce Manager will report a feed 95 percent active while the share of SKUs actually eligible to serve sits somewhere between a third and two thirds of the range: variants collapsed into one item, availability lagging a 24-hour fetch on stock that turns over in six, prices carrying Swedish VAT into a Danish feed. We start by measuring served eligibility against catalogue truth taken from your commerce platform's own export, SKU by SKU. The gap between those two numbers is usually the entire conversation. What this work touches: The feed source first. Centra, Shopify, Litium, Norce, or the XML somebody wrote in 2019 and left behind. We read the export, not the dashboard. Then Meta Commerce Manager, its catalogue diagnostics, feed rules and set-level overrides, TikTok Catalog Manager, and Pinterest Catalogs where the range justifies a third channel. Identity fields are fixed before anything else: retailer_id stability across PIM re-indexes, item_group_id so a shoe in nine sizes is one product rather than nine competing against each other, GTIN and MPN where the platform uses them for matching. Availability is the second pass. Scheduled fetch is the wrong mechanism for a range that sells through in days. We move stock and price onto the Batch API or a webhook-driven push and leave scheduled fetch as the reconciliation path rather than the source of truth. Then segmentation. The custom_label fields are the only real lever you own inside Advantage+ catalogue ads, and they are usually empty or filled with a merchandiser's season codes. We rebuild them around what bidding needs: contribution margin band, stock cover in weeks, return rate band, lifecycle stage. Sets are built from those labels, not from the website's navigation. Multi-market Nordics: One catalogue, country feed overrides for SE, NO, DK and FI covering currency, VAT treatment, delivery promise, and language on title and description. We verify the override is applied against the ad account's country setting rather than assumed. The failure mode here is silent and surfaces as a Norwegian shopper reading a Swedish price with 25 percent VAT already inside it. Not included: We do not build or host feeds and we do not resell a feed management tool. If your commerce platform cannot export what the work requires, we specify exactly what your engineers must change and hand that over. No Google Merchant Center work: the specifications differ enough that pretending otherwise is how a company ends up with one mediocre feed serving both. No creative production for catalogue ads, though we will write the overlay and frame rules and hand them to whoever makes them. Who this is not for: Ranges under a few hundred SKUs, where a hand-maintained CSV is genuinely fine and this engagement would be an expensive way of confirming that. And anyone who wants the diagnosis without acting on it: roughly a third of what we find is fixable inside the platform, and the rest becomes a ticket for your commerce team. If that ticket has nowhere to go, the engagement will not pay for itself and we would rather say so at the start. What you receive: A field-level defect list ordered by revenue exposure with the affected SKUs listed under each line. The corrected catalogue and set structure built in your own Business Manager. A written feed specification your engineers can implement against. And a monitoring sheet with the four numbers worth checking weekly: served eligibility, availability lag in hours, variant collapse rate, and rejected items by reason. We walk your team through it once, recorded.