ERP Data Migration Checklist: 40 Steps From Legacy Extract to Cutover Sign-Off
Forty steps, each with an owner, a deliverable and an exit criterion. It covers how many mock loads to run, which open balances must tie to zero and which history belongs in an archive rather than the new ERP.
Who Owns Each Step and Why Exit Criteria Matter
The cutover weekend can end two ways. Finance signs the opening balances, or people are still arguing at 3 a.m. about a receivables gap nobody can explain. Which one you get is mostly settled months earlier. This ERP data migration checklist breaks the work into 40 steps across seven phases. Every step has one owner, one deliverable and one exit criterion. If a step's exit criterion isn't met, the next phase doesn't start. That one rule matters more than any tool you buy.
The checklist uses six roles:
- Migration lead: runs the program. This is usually the systems integrator's data architect or an internal data manager.
- Data owners: one business person per object (customers, vendors, items, chart of accounts, assets).
- Financial controller: owns control totals and reconciliations.
- Legacy IT: owns the source systems and the extracts.
- SI functional consultants: own the mapping and the load programs.
- Steering committee: approves scope and makes the go/no-go call.
Every object should have a business person as its data owner, not someone from IT. IT can move a vendor record. Only AP knows whether it's the right vendor with the right bank account.
Scope and the Migrate-or-Archive Decision (Steps 1–6)
- 1. Inventory every data object. Owner: migration lead. Deliverable: object register. Exit: approved by the steering committee.
- 2. Classify each object as master data, open transaction or history. Owner: migration lead with data owners. Deliverable: a classification column in the register. Exit: every object is classified.
- 3. Set the history cutoff. Owner: controller. Deliverable: written history policy. Exit: signed by the CFO.
- 4. Confirm retention obligations in every country you operate in. Owner: tax and legal. Deliverable: retention matrix by legal entity. Exit: reviewed by a local adviser in each jurisdiction.
- 5. Choose the archive approach. Owner: IT. Deliverable: archive design, which can be a read-only database, a data lake, or the legacy system kept on a read-only license. Exit: the external auditor confirms they can reach the data.
- 6. Freeze scope. Owner: steering committee. Deliverable: baselined register. Exit: any later change goes through change control.
Here's the default we'd recommend for most companies. Migrate active masters (anything used in the last 18–24 months), all open items, and GL balances summarized by period for the current and prior fiscal years so comparative reporting still works. Send transaction-level history to the archive. Each extra year of detailed history makes every mock load longer and gives you more variances to chase, and it rarely buys you anything an archive query couldn't.
Your archive design depends on retention rules, and those differ a lot from country to country. In Germany, commercial and tax records generally have to be kept for eight to ten years depending on the record type, and tax auditors expect machine-readable access. UK tax records are typically kept for around six years. Indian company law requires books of account for eight financial years. Most other jurisdictions fall somewhere between five and ten years. Data-protection law pushes in the opposite direction: the EU's GDPR, India's DPDP Act and Brazil's LGPD all discourage holding personal data you no longer need. These rules change, so have a qualified tax adviser confirm each entity's obligations. A read-only archive typically costs $20,000–$100,000 to set up, plus hosting. That's a typical 2026 market range and varies by contract and region, so get vendor quotes.
Legacy Extract and Data Profiling (Steps 7–12)
- 7. Build scripted, repeatable extracts. Owner: legacy IT. Deliverable: extract scripts, with no manual spreadsheet exports. Exit: two runs against unchanged data return identical counts.
- 8. Profile every object. Owner: migration lead. Deliverable: a report on nulls, invalid formats, orphan records and outliers. Exit: reviewed by each data owner.
- 9. Identify duplicates. Owner: data owners. Deliverable: duplicate list with survivorship rules. Exit: rules signed.
- 10. Flag inactive records. Owner: data owners. Deliverable: an inactive list based on a rule such as no activity in 24 months and a zero balance. Exit: approved for exclusion.
- 11. Capture control totals at extract. Owner: controller. Deliverable: record counts and value totals for each object. Exit: totals tie to the standard legacy reports.
- 12. Secure the extract. Owner: IT security. Deliverable: an encrypted staging area with access logging. Exit: personal and bank data are masked in all non-production environments.
Step 7 is the one teams most often shortcut, and they shouldn't. When an extract is a spreadsheet someone exported by hand, nobody can tell whether a mock 3 variance came from the data or from the person doing the export.
ERP Data Cleansing and Field Mapping (Steps 13–19)
- 13. Write the field-level mapping spec. Owner: SI functional consultant. Deliverable: source-to-target mapping with transformation rules. Exit: the data owner signs off each object.
- 14. Build code crosswalks. Owner: controller and operations lead. Deliverable: crosswalk tables for chart of accounts, units of measure, payment terms and warehouse locations. Exit: no unmapped legacy values remain.
- 15. Decide where cleansing happens. Owner: migration lead. Deliverable: decision log. Exit: agreed by the data owners.
- 16. Complete mandatory fields. Owner: data owners. Deliverable: populated tax and e-invoicing identifiers (EU VAT numbers, India GSTINs, Mexico RFCs, Saudi VAT registration numbers), plus HS codes and Incoterms wherever trade requires them. Exit: no blanks in mandatory target fields.
- 17. Verify vendor bank details. Owner: AP lead. Deliverable: bank records confirmed by calling back a known contact. Exit: every active vendor verified. Changed bank details slip through easily while records are being reloaded and nobody is watching closely, and fraudsters know that.
- 18. Set cleansing KPIs. Owner: migration lead. Deliverable: error-rate dashboard by object. Exit: every object is below its agreed threshold before mock 3.
- 19. Put master data under change control. Owner: data owners. Deliverable: dual-maintenance procedure. Exit: in force from mock 3 onward.
On step 15: if legacy stays live for several more months, clean the data in the source system. Fixes made there carry into every later extract automatically. Clean in staging only when the schedule leaves no time for source fixes, and accept that the transformation rules then have to be rerun and retested at every mock.
How Many Mock Data Loads to Plan (Steps 20–26)
A single-entity, mid-market project should plan three mock loads plus a dress rehearsal. A multi-entity, multi-country rollout should budget four mocks, because every additional currency, tax regime and intercompany flow brings its own variances. Leave three to six weeks between mocks so defects actually get fixed. On a 9–12 month implementation, mock 1 usually comes around month five or six.
The thresholds below are suggested exit criteria. Agree your own with the steering committee before mock 1.
| Load | Data scope | Exit criterion |
|---|---|---|
| Mock 1 | Masters at sample volume | Load programs run; at least 80% of records load |
| Mock 2 | All masters plus open items | At least 95% load; first reconciliation complete |
| Mock 3 | Full production volume | At least 99.5% of masters load; balances tie with zero unexplained variance; run time measured |
| Dress rehearsal | Full cutover runbook | Completes within 80% of the cutover window |
- 20. Script the environment refresh. Owner: SI technical lead. Exit: the target environment can be reset in under a day.
- 21. Run mock 1. Owner: migration lead. Deliverable: load log. Exit: per the table.
- 22. Run mock 2. Owner: migration lead. Exit: per the table.
- 23. Run mock 3. Owner: migration lead. Exit: per the table.
- 24. Business validation of samples. Owner: data owners. Deliverable: 25–50 records per object checked on screen. Exit: signed validation sheet.
- 25. Fix defects. Owner: SI. Deliverable: defect log. Exit: no open severity-1 or severity-2 defects.
- 26. Run the dress rehearsal. Owner: migration lead. Exit: per the table.
Open Balances to Reconcile Before Anyone Signs (Steps 27–32)
| Balance | Reconcile against | Tolerance |
|---|---|---|
| GL trial balance by entity, account and currency | Legacy closing trial balance | Zero |
| AR and AP open items | Legacy aged reports | Zero on value; item counts match |
| Inventory quantity and value by site, bin and lot | Count results and valuation report | Zero on quantity; agreed rounding on value |
| Fixed asset cost and accumulated depreciation | Asset register | Zero |
| Open POs, open SOs, goods received not invoiced | Legacy open-order and GRNI reports | Zero |
| Bank, tax and intercompany | Statements, tax returns, counterparty balances | Zero; intercompany nets to zero |
- 27. Agree reconciliation templates. Owner: controller. Exit: the auditor has reviewed them.
- 28. Reconcile the GL. Owner: controller. Exit: per the table.
- 29. Reconcile subledgers to the GL. Owner: AR, AP and asset leads. Exit: each subledger equals its control account.
- 30. Reconcile inventory and GRNI. Owner: warehouse lead with the cost accountant. Exit: per the table, including stock held at 3PLs.
- 31. Reconcile tax and intercompany. Owner: tax lead. Exit: each jurisdiction ties to its latest filing position.
- 32. Document every variance. Owner: controller. Deliverable: variance log with root causes. Exit: no line reads "unexplained."
Here's a hypothetical example. A three-entity distributor gets to mock 3 with an $18,400 AR variance. The cause is 37 foreign-currency invoices that were loaded at the cutover rate, when legacy had revalued them in a different period. The fix is to load them at the original document rate and post the revaluation separately. A sample-volume mock would never have turned that up, which is why mock 3 has to run at full volume.
ERP Cutover Plan: Freeze, Load, Verify (Steps 33–38)
- 33. Publish the cutover runbook. Owner: migration lead. Deliverable: hour-by-hour task list with a named person on every line. Exit: proven in the dress rehearsal.
- 34. Set the transaction freeze. Owner: operations lead. Deliverable: stop times for shipping, receiving and invoicing, communicated to warehouses, 3PLs and carriers. Exit: written confirmation from every site.
- 35. Run the final extract with control totals. Owner: legacy IT and controller. Exit: totals tie to the closed legacy period.
- 36. Load production. Owner: SI. Deliverable: load log. Exit: rejection rates no worse than mock 3.
- 37. Reconcile after the load. Owner: controller. Exit: every line in the steps 27–32 table ties.
- 38. Hold the go/no-go. Owner: steering committee. Deliverable: decision record measured against criteria agreed before cutover. Exit: go, or rollback triggered.
Cut over at a period end, once the legacy period is closed. One month-end close gets squeezed, but you open on clean balances and avoid weeks of separating half a month of activity between two systems.
Sign-Off, Hypercare and the Red Flags That Should Stop the Program (Steps 39–40)
- 39. Get formal sign-off. Owner: CFO. Deliverable: a sign-off pack containing every reconciliation and the variance log. Exit: signed by the CFO, the controller and every data owner.
- 40. Run hypercare, then decommission legacy. Owner: migration lead and IT. Deliverable: four to eight weeks of daily defect triage, with legacy switched to read-only. Exit: decommission only after the first period close, the year-end audit and a successful test restore from the archive.
Halt the program if you see any of these:
- No full-volume mock before the dress rehearsal.
- Every data owner sits in IT.
- Variances described as "within tolerance" without a root cause.
- Manual spreadsheet fixes made after mock 3.
One rule outranks all the others. If mock 3 doesn't reconcile with zero unexplained variance, move the go-live date and leave the tolerance where it is.
Frequently asked questions
How long does ERP data migration take within the overall project?
On a typical 9–12 month mid-market implementation, the data workstream usually runs for five to seven months. Extracts and profiling start early, and cutover comes at the end. Multi-country rollouts take longer because each entity brings its own tax, currency and retention requirements. Leave at least three to six weeks between mock loads so defects actually get fixed rather than carried forward.
Should we migrate all historical transactions into the new ERP?
Usually not. Migrate active masters, all open items, and GL balances summarized by period for the current and prior fiscal years. Put transaction-level history in a read-only archive that auditors and tax authorities can query. Retention periods generally range from about five to ten years depending on the country and record type, so confirm each entity's obligations with a qualified tax adviser.
What does ERP data migration typically cost?
For planning, assume data migration will take roughly 10–20% of implementation services fees. Then add tooling and the archive; a read-only archive often costs $20,000–$100,000 to set up. These are typical 2026 market ranges that vary widely by contract, data volume and region, so get quotes from your integrator and archive vendors. Cost is driven mainly by how many legal entities you have, how much history is in scope and how dirty the legacy data is.
Who should sign off the migrated opening balances?
The CFO signs last. Before that, the financial controller has to sign every reconciliation and each data owner has to sign their object. Ask your external auditor to review the reconciliation templates before mock 3, not after go-live, so the sign-off pack is already in a form they'll accept.
Written by the Port Cloud Pro editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.