How to Clean and Migrate a Complex Product Catalog

1. Audit: learn what you actually have

Export everything and measure it: total records, duplicate and near-duplicate rates, attribute fill rates, image coverage, variant integrity. This is not busywork – the audit decides the project's shape. A catalog with 4% duplicates is a cleanup task; one with 30% is a rebuild decision.

2. Dedupe: decide what "same product" means

Duplicates are rarely exact. Define matching rules (SKU plus key attributes, or fuzzy title matching with manual review above a threshold) and – critically – decide survivorship: when two records merge, which title, image set, and description wins. Document the rules; you will need them when someone asks why a product changed.

3. Normalize: one standard for everything

Pick canonical formats for naming, units, attributes, and categories, then convert the catalog to them. This is tedious and it is the highest-value step: consistent data migrates cleanly, imports correctly, and feeds Google Shopping without drama. Inconsistent data does none of those things.

4. Map: every field, deliberately

Build the field map between source and target platforms, including custom attributes and edge cases. Pay special attention to variants (every platform models them differently), SEO fields (titles, descriptions, URLs), and anything the target platform simply does not have – those need a conscious decision, not a silent drop.

5. Migrate in stages, validate each one

Move a representative slice first – a few hundred products across all your edge cases – and validate imports, display, variants, and links. Only then run the full migration. Keep the old catalog intact and tested rollback ready until reconciliation passes.

6. Verify: prove it worked

Reconcile record counts, spot-check listings, verify URL redirects, and confirm feeds and integrations see the new data correctly. The project is done when the numbers prove it, not when the import finishes.

Related: Feed Diagnostics · Diagnosing feed errors