Data migration is the part of an ERP program that everyone underestimates and every go-live depends on. Searches like “ERP data migration best practices,” “ERP data migration strategy” and “ERP data migration plan template” reflect that worry. For manufacturers the challenge is specific: items, bills of material, routings and costs have to be right on day one, or plants cannot build and finance cannot close.
- Scope the data by what the business needs on day one, not by what the old system holds. Most history can stay in an archive.
- Manufacturing master data (items, BOMs, routings, costs) needs business owners in engineering and finance, not just IT.
- Plan at least three mock loads, each reconciled and signed off, before cutover.
- Name a data migration lead early. It is one of the hardest roles to fill late in a program.
Why manufacturing data is harder
A distributor can go live with clean customers, items and open orders. A manufacturer also needs every bill of material, routing, work center, cost and quality specification to be right, because the new system will plan, cost and report production from them. Errors that were tolerated in the old system, through workarounds people stopped noticing, show up immediately as wrong MRP suggestions, wrong costs or blocked production orders.
A data migration plan in seven steps
- Define scope by object. List every data object (customers, suppliers, items, BOMs, routings, work centers, costs, open orders, on-hand inventory, fixed assets, open AR and AP) and decide what moves, what is archived, and how much history to bring.
- Assign business owners. Each object needs an owner who defines “correct” and signs the reconciliation. For BOMs and routings that is plant engineering; for costs it is finance.
- Profile and cleanse early. Find duplicates, obsolete items, inconsistent units of measure and incomplete BOMs while design is still underway. Cleansing at the source is cheaper than fixing at load time.
- Map to the new design. Field mappings follow design decisions, so data and functional leads must work together. Changes to item types, units of measure or costing methods ripple through every object.
- Build repeatable loads. Use the target system's standard import tools and scripts that can be rerun. Manual spreadsheets do not survive three mock loads.
- Run mock loads and reconcile. At least three: an early technical load, a full load for testing, and a dress rehearsal timed like the real cutover. Reconcile quantities and values to the legacy system every time.
- Freeze and cut over. Agree data freeze dates with plants and purchasing, then load and reconcile the final balances under the cutover runbook.
If a bill of material is wrong in the old system, it will be wrong in the new one, faster.
Reconciliation: how you prove it worked
- Record counts by object, with explained differences.
- Inventory quantity and value by plant, warehouse and item, tied to the general ledger.
- Open order quantities and values for sales, purchasing and production.
- Sample-based checks on BOMs and routings by plant engineers, including a test build of selected products.
- Open AR and AP tied to subledgers and agreed with finance.
The roles that keep data on schedule
Most programs need a data migration lead who owns the plan and the reconciliation, one or more data analysts or developers who build the loads, and business data owners released from part of their day job. In Oracle Fusion that means FBDI and API loads; in SAP S/4HANA, the Migration Cockpit or data services. If the program is moving off SAP ECC or Oracle EBS, see our guides to SAP ECC end of support options and Oracle EBS to Fusion migration.
Maxary places ERP data migration leads and analysts with manufacturing experience. See what we staff.
Frequently asked questions
What are ERP data migration best practices?
Scope data by what is needed on day one, assign business owners for each object, cleanse at the source early, map to the new design, use repeatable load tools, run at least three reconciled mock loads, and freeze data before cutover.
How many mock data loads should an ERP project run?
At least three: an early technical load, a full load to support testing, and a timed dress rehearsal of the cutover. Each should be reconciled and signed off.
Who should own ERP data migration?
A data migration lead owns the plan and reconciliation, but business owners must define and sign off each object, for example plant engineering for BOMs and routings and finance for costs and balances.
How much history should move to the new ERP?
Usually only open transactions and balances, plus master data. Most history can be archived and kept available for reporting and audit without being converted.