§ 01 — why migrations stall
One long migration is three stalled ones.
The default plan — 'extract from Excel, map to PIM, load' — fails because it compresses work that must happen sequentially into one phase. The data model isn't finalised when migration starts, so attributes keep being renamed. Enrichment hasn't happened, so operators load garbage. Cutover criteria aren't written, so nobody can call 'done'.
The fix is five phases, each with an explicit exit condition. Below.
§ 02 — phase 1
Inventory & taxonomy (2–3 weeks).
- Catalogue every attribute across every source spreadsheet. Dedupe by meaning, not by column name — 'weight' and 'net_weight_g' are the same attribute.
- Group into attribute families (regulated, logistics, marketing, media).
- Kill 30–50% of legacy attributes. Most spreadsheets accumulate attributes nobody owns, nobody updates, nobody reads.
- Exit: a canonical attribute list agreed by data steward, merch lead, and one downstream consumer (e.g. channel manager).
§ 03 — phase 2
Model in PIM (2–4 weeks).
- Implement families, variants, locales, channel profiles in PIM — empty of data.
- Build required-attribute rules per channel. This is the completeness score schema.
- Define the three-tier workflow: draft → review → publish.
- Exit: a handful of test SKUs authored end-to-end by a real merchandiser, not a consultant.
§ 04 — phase 3
Pilot migration (3–6 weeks).
- Pick one product family with real downstream consumers. Not the bestsellers — the representative SKUs.
- Migrate, enrich, publish to one channel. Measure completeness before and after.
- Fix every data-quality issue at the source. If the legacy spreadsheet has bad data, clean it before it enters PIM.
- Exit: the pilot family is live on the target channel at your completeness threshold, and authoring time per SKU is measured.
§ 05 — phase 4
Bulk migration (6–12 weeks).
- Run the pilot playbook, family by family. Parallelise 2–4 families at a time.
- Enforce a soft freeze on the legacy spreadsheets mid-migration: updates allowed only in PIM for families already migrated.
- Track throughput (SKUs/week) and completeness as KPIs in the steering committee.
- Exit: all families migrated, legacy spreadsheets frozen, dual-write disabled.
§ 06 — phase 5
Cutover & hypercare (4–8 weeks).
- Kill the legacy spreadsheets permanently. Archive, don't leave editable.
- Run parallel publishes to channels for 2 weeks: old and new. Reconcile.
- Flip the switch; run hypercare with SI partner on-site / on-call.
- Exit: 4 weeks post-cutover with no revert, no data rollback, no dual-source attribute updates.
The test of a finished migrationIf your merchandising team can still update a SKU in the old spreadsheet, the migration isn't finished — regardless of what the steering committee slides say.
Next steps on pimvendors.com