Skip to content

Business Systems

ERP Data Migration: How to Move Clean Data Without Downtime

Bad data makes a good ERP look bad. Learn what to migrate, how to clean and map it, how to test the move and how to cut over safely.

By the CodeElectron engineering teamUpdated 6 min read
Illustration of a central hub connected to several data nodes by glowing lines

Key takeaways

  • Decide what to migrate: master data and open items are essential, long history is usually optional.
  • Clean first. Remove duplicates, fix formats and agree one naming standard before moving anything.
  • Run at least two test migrations and reconcile totals, balances and stock before the real one.
  • Plan the cut-over: freeze changes, migrate, verify, then switch, with a fallback if checks fail.

Why data migration decides ERP success

Users judge an ERP by what they see in the first week: do the customer balances match, is the stock correct, are the prices right? If the migrated data is wrong, trust is lost quickly, however good the software is. Migration is not a final technical task. It is a project of its own, with owners, tests and deadlines, and it should start early.

What to migrate and what to leave behind

Data typeExamplesUsually
Master dataCustomers, suppliers, products, price lists, chart of accountsMigrate after cleaning
Open itemsUnpaid invoices, open orders, current stock, open balancesMigrate
Opening balancesAccount balances at the cut-over dateMigrate and reconcile
HistoryPast invoices and transactionsMigrate a limited period, archive the rest
AttachmentsScanned documents, contractsOnly if needed, or keep in an archive

Clean before you move

  • Find and merge duplicate customers, suppliers and products.
  • Standardise names, units, currencies and codes.
  • Fill missing required fields such as tax numbers and addresses.
  • Remove inactive records that nobody needs.
  • Decide who is responsible for each data set and approves its quality.

Map fields and rules

For each source field, decide the target field and any transformation: split a combined name, convert units, map old categories to new ones. Write the mapping down so it can be reviewed by the business and repeated exactly in every test.

Test the migration more than once

  1. 1Run a first migration into a test system and review the problems.
  2. 2Fix the mappings and clean the data again.
  3. 3Run a second full migration and reconcile: record counts, customer and supplier balances, stock quantities and values, account totals.
  4. 4Let key users check real records they know well.
  5. 5Repeat until the results match and are signed off.

Plan the cut-over

  • Choose a quiet period, such as a month end or a weekend.
  • Agree a freeze: no changes in the old system after a set time, or capture changes for later.
  • Run the final migration and the reconciliation checks.
  • Switch users over only after the checks pass.
  • Keep the old system available read-only for a while, with a rollback plan.

A worked example

A distributor moves from spreadsheets to a new ERP. Test migration one reveals three spellings of the same customer, products without units and negative stock. After cleaning, test migration two matches stock quantity and value to the physical count. The final migration happens over a weekend, and on Monday the team already trusts the numbers. This is an illustrative scenario, not a client story, but the lesson is real: the dirtier the data, the earlier you must start.

Next step

Data is one step of a bigger plan. Read the full ERP implementation plan, explore our ERP and business systems service, or contact us to review your data before a move.

Frequently asked questions

How much history should I migrate to a new ERP?

Open items and opening balances are essential. Past transactions are usually migrated for a limited period and the rest archived, because old history adds risk and rarely helps daily work.

Can data migration be done without downtime?

Often the downtime can be kept short by testing migrations in advance, choosing a quiet period and preparing the final run. Plan a freeze window and a fallback.

Who is responsible for data quality?

The business owns its data. The project team provides tools and checks, but each data set needs a named owner who approves it.

Have a question we have not covered?

Ask the team directly. We are happy to talk through your situation.