A useful guide does not promise certainty. It gives you better questions, a shared vocabulary and a clear place to start.
Audit and map
Products, variants, customers, orders and subscriptions mapped field by field before any import runs.
Dry run and reconcile
A full test import into staging, reconciled against the source system and signed off by whoever owns the numbers.
Redirects
Every indexed URL mapped to a destination, chains collapsed, sitemaps regenerated for submission at launch.
Cutover
A timed runbook with owners, checkpoints, comms and a rollback path if any check fails.
A migration is a data project wearing a design project’s clothes. Nothing should go live until every item here is signed off by the person who owns the numbers.
- Field-level mapping document agreed in writing
- Explicit list of what deliberately does not migrate
- Fields the new platform cannot hold, with an archive plan
- Full dry run imported into a staging store
- Record counts reconciled per entity against the source
- Twenty historic orders spot-checked end to end, including refunds
- Subscription state and next billing dates verified
- Every indexed URL mapped to a destination, chains collapsed
- Legacy parameter and filter URLs handled
- Payments, tax, shipping, ERP and email reconnected and tested
- Cutover window chosen outside peak trading
- Runbook with an owner and deadline against every step
- Live-key test transaction completed before announcing
- Documented rollback trigger and the person who calls it
14 items · no email required
Where teams go wrong with this
Four failure patterns we have watched play out often enough to name. None of them is caused by a lack of effort; all of them are caused by sequencing.
Migrating during peak
The launch itself may go smoothly. What you lose is the buffer to fix what the first fortnight of real traffic reveals. Move in a quiet quarter and use peak to trade.
Reconciling after the fact
A dry-run import that nobody reconciles against the source system is a rehearsal with the ending missing. Whoever owns the numbers should sign off the test import before the real one is scheduled.
Redirect chains left in place
Old URL to interim URL to new URL leaks authority at every hop and is slower for every visitor. Collapse chains to a single redirect and regenerate sitemaps for submission at launch.
Integrations tested in isolation
Tax, shipping, ERP and email each work alone and fail together. Test the full order lifecycle end to end, including a refund and a subscription renewal.
Questions this guide raises
Temporarily and mildly if the redirect map is complete, the URL structure is deliberate and the content survives intact. Materially and lastingly if any of those three are handled in the final week. The variable is preparation, not the platform.
Eight to sixteen weeks for most mid-sized catalogues, with the data mapping and reconciliation taking longer than the storefront build. Subscriptions, complex tax or bespoke ERP integration push it further.
Sometimes — running the new storefront for one region or one catalogue segment first is a genuine risk reducer. Splitting checkout across two platforms is not; that one has never gone well for anyone we know.
- Website Redesign Planning GuideThe questions to settle before design starts.
- App Launch Readiness ChecklistA practical worksheet for the fortnight before you submit to the stores.
- Web Accessibility Audit TemplateA clear, team-friendly way to turn WCAG findings into action.