How to Validate SKUs Before They Enter Your Catalog

A malformed SKU is cheap to catch the moment it arrives and expensive to unwind once it’s propagated into orders, pricing rules, and three months of sales history. Most catalogs don’t check at the door. A supplier feed lands, gets mapped, gets imported — and whatever shape the SKUs are in becomes your problem downstream, usually discovered when a barcode scan fails at the warehouse or two clearly different products share an identifier.

What “validating a SKU” actually checks

Format correctness. Does it match the expected structure — length, character set, check-digit logic for GTIN-based codes? This catches transcription errors and malformed exports before they’re trusted anywhere. See common barcode errors for the specific failure patterns.

Uniqueness against the existing catalog. Is this SKU already assigned to a different product? Supplier renumbering, reused codes after a product is discontinued, and cross-supplier collisions all produce this — silently, until two products share an identifier and one becomes invisible in searches for the other.

Consistency with claimed identity. If a row claims a specific GTIN, does that GTIN’s manufacturer prefix match the stated brand? Mismatches here usually indicate a transcription error or a genuinely different product mislabeled.

Non-duplication against the incoming batch itself. Supplier files sometimes contain the same product twice within one export — a validation pass needs to catch this before it’s compounded with the “existing catalog” check above.

The workflow

1. 1
   Format check every incoming identifier

   Validate each incoming SKU, GTIN, and MPN against its expected structure. Reject or flag anything malformed before it goes further.

2. 2
   Match against the existing catalog

   Use the identifier with a confidence score. An exact match updates the existing product; a partial or fuzzy match queues for review rather than silently creating a new record.

3. 3
   Cross-check brand and prefix consistency

   Where the identifier encodes manufacturer information, compare that signal against the stated brand or manufacturer before trusting the row.

4. 4
   Flag, don't silently drop

   Anything failing validation should route to a short human review queue with the reason attached, not disappear or get force-imported anyway.

Done at the point of ingestion, this is a filter that runs in seconds per file. Done after the fact — reconstructing which of ten thousand SKUs are malformed once they’re already live in orders — it’s a forensic project.

Related reading

FAQ