Connecting marketplaces multiplies your catalog: every product becomes three or four listings, every attribute feeds three or four search engines, and every defect gets copied at machine speed. Cleanup BEFORE connection is a week of unglamorous work; cleanup after is a quarter of whack-a-mole. Five passes, in order.
Pass 1: SKU normalization
The join key for everything downstream, matching, sync, forecasting, is the SKU. The pass: every product and every variant gets one canonical identifier per the naming convention, blanks filled, near-duplicates (SKU-1 vs sku_1) unified, and retired SKUs never recycled. This pass alone converts marketplace connection from archaeology into an afternoon.
Pass 2: duplicate execution
Find and merge duplicates: same product under two listings, variants listed as standalone products, test products that escaped. Keep the copy with history, zero and retire the rest, and reconcile the freed stock into one honest count, connecting a catalog with internal duplicates gives every marketplace a copy of your confusion.
Pass 3: variant structure
Marketplaces are opinionated about parent-child structure (Amazon variation families, eBay variations, Etsy attributes). The pass: consistent variant dimensions per family (color-then-size, always), child SKUs everywhere, and no families mixing variation types. Broken variant structure is the top source of listing-creation errors when tools push to marketplaces.
Pass 4: data quality
Every channel will render what you have: titles present and honest, descriptions real, weights and dimensions filled (shipping calculators read them), categories assigned, GTINs where your products carry them. The bar is not marketing polish, it is completeness: blank fields become blank listings at scale, the data-quality pass has its own checklist.
Pass 5: asset ownership
The inkFrog shutdown made this pass famous: listings that referenced a vendor’s hosted images and templates broke the day its servers died, the full episode. The pass: images in storage YOU control (your store’s media library counts), descriptions that render without third-party CSS, and an export of everything, so no single vendor’s disappearance reaches into your live listings.
The readiness test
One export, one criterion: dump the catalog to a spreadsheet, and check whether a stranger could reconstruct your store from it, unique SKUs, one row per variant, complete fields, image URLs you own. If it round-trips without hand-editing, connection day is an afternoon: store as source of truth, marketplaces matched by SKU, rules set, the five-scenario sync test, live.
Common questions
How long does cleanup actually take?
A few evenings for a few hundred disciplined SKUs; a real week for years of accretion. Either way it is bounded, the post-connection version of the same mess is not.
Should I clean up in my store or in a spreadsheet?
The store is the source of truth, so fixes land there; the spreadsheet is the audit surface, and its retirement plan starts right after.
Can I connect first and clean gradually?
You can, and every defect will be live on more surfaces while you do. Clean the SKUs and duplicates first at minimum; those two passes are what the connection mechanics depend on.
What about products I do not plan to cross-list?
Clean them anyway, catalog hygiene is per-catalog, not per-channel, and channel selection per product is easier from a clean base.
Clean in, honest everywhere
Connect a clean catalog and let one pool keep every channel true, from $49/month with unlimited orders. See pricing.