ERP testing for manufacturers: what must you prove before go-live?
The test levels and the evidence each must produce, a production-flow script from customer order to ledger, checks for migrated stock and open work, and go/no-go criteria for a manufacturing ERP.
- 01Configure
- 02Integrate
- 03Run the flow
- 04Accept
- 05Cut over
ERP testing for a manufacturer must prove that the configured system can run a real order from demand to cost, not only that each screen works. Test in layers: configuration, integrations, the end-to-end production flow, user acceptance on the shop floor, migrated data and a rehearsed cutover. Go live only when a mock cutover reconciles stock, work in progress and the ledger, trained users complete their scripts unaided, and every open defect has an agreed workaround.
The vendor demonstration worked and the configuration is signed off. Then, in the first week live, an open work order ignores a revised bill of materials, operators cannot record scrap on the floor terminals, and finance cannot explain the work-in-progress balance. None of those failures is unusual. Each would have appeared in a test that followed one real product through the factory with real data and the people who will use the system. This guide sets out that test.
What does ERP testing need to prove for a manufacturer?
ERP testing checks the configured system, its data and its connections against the way the business will actually operate, before it becomes the system of record. In a factory the risk sits where planning, the shop floor and finance meet. A purchasing screen can pass on its own while the material plan it feeds is wrong. A work order can close while the cost it posts cannot be explained.
Implementation testing is different from the proof you ran when choosing software. A selection demonstration shows what the product can do with the vendor’s data. Implementation testing shows what your configuration does with your items, bills of materials, routings, stock and people. If you are still choosing, start with the manufacturing ERP software comparison and its proof script.
| Test level | Question it answers | Who runs it | Exit evidence |
|---|---|---|---|
| Configuration testing | Does each setting behave as the design says it should? | Implementation team | Each setting checked against its design note; defects logged with an owner. |
| Integration testing (SIT) | Do records move to and from connected systems correctly, including when something fails? | Implementation and integration team | Orders, stock, shipments and postings arrive once and in the right state; rejected messages are visible. |
| End-to-end flow testing | Can one order travel from demand to cost? | Process owners with the implementation team | The production-flow script passes with reconciled quantities and values. |
| User acceptance testing (UAT) | Can the people who will use the system do their work without help? | Planners, buyers, supervisors, operators and finance | Signed scripts, run without consultant narration, on the devices people will use. |
| Data migration testing | Is the loaded data complete and correct? | Data owners in each function | Items, BOMs, stock, open orders and balances reconcile to the source. |
| Volume testing | Does the system cope with a busy day? | Technical team | Planning runs, scanning and posting complete within agreed times at peak volume. |
| Cutover rehearsal | Can the business switch systems within the planned window? | Project lead with every function | A timed mock cutover with signed reconciliation and a tested fallback. |
| Regression testing | Did a fix or update break something that had passed? | Implementation team | Core scripts rerun after every change before it reaches users. |
Integration testing (often called system integration testing, or SIT) and user acceptance testing (UAT) are the two levels most often squeezed when a project runs late. In manufacturing they are also the two that catch the most expensive problems: an order arriving twice from a storefront, or an operator who cannot record scrap at the machine and writes it on paper instead.
Which production flow should the test follow?
Choose one representative product family and follow it from the customer order to the ledger. The illustrative example below is a made-to-order assembly with a subassembly made to stock, two purchased components, an outside coating operation and a final inspection. A real test should use the business’s own product, with the exceptions that actually happen in its factory.
- 01Accept the order and promise a dateSales or commerce, checked by the ERP
Order line, configured revision, quantity and requested date
- 02Plan material and capacityERP planning (MRP)
Planned purchase and work orders, shortages and capacity load
- 03Release and run the work orderProduction supervisor
Material issued by lot, labour and machine time, output and scrap
- 04Inspect and releaseQuality
Inspection result, hold or release, and any non-conformance
- 05Receive finished stock and dispatchWarehouse
Finished lot or serial number and shipped quantity by line
- 06Cost, invoice and closeFinance
Actual cost, variances, invoice and ledger postings
If orders arrive from a B2B portal or storefront, or stock moves through a separate warehouse system, include those systems in the same run. The handoff rules in the ecommerce ERP integration guide become test cases: one order created once, stock published from one owner, and dispatch reported by line.
What does a production-flow test script include?
Run the script with migrated or production-like data, not the demonstration company. Each step is pass or fail, and each one includes an exception on purpose, because normal orders rarely expose a weak design. In the illustrative case the customer orders 10 units, one unit is scrapped and one fails final inspection.
| Step | Action | Pass when | Fail signs |
|---|---|---|---|
| 1 | Enter a sales order for 10 configured units with a date that current capacity cannot meet. | The ERP shows the constraint and proposes a date that sales can explain to the customer. | The requested date is accepted without checking material or capacity. |
| 2 | Release revision C of the bill of materials while a revision B work order is still open. | The open order keeps revision B, new orders use C, and planners can see which order uses which. | The open order silently switches revision. |
| 3 | Run material planning with one purchased component short. | A planned purchase order and a shortage message point back to the demand line. | Planned orders appear that nobody can trace to demand. |
| 4 | Substitute an approved alternate component. | The substitution is recorded on the work order and costed at the alternate. | The substitute is issued off-system and the record shows the original part. |
| 5 | Release the work order and issue material by lot. | Stock moves to work in progress (WIP) with the lots recorded. | Automatic consumption at completion (backflushing) records standard quantities and no lots where traceability is required. |
| 6 | Record labour and machine time from the shop-floor terminal. | The operator completes the task in the agreed number of steps on the real device. | A supervisor re-keys the shift from paper. |
| 7 | Send the subassembly to the coating supplier and receive it back. | Material at the supplier stays visible, and the service cost and receipt land on the work order. | Stock disappears between the purchase order and the receipt. |
| 8 | Complete 9 units and scrap 1. | Scrap is posted with a reason, and planning sees a shortfall of one unit against the customer line. | The work order shows 10 complete. |
| 9 | Fail 1 unit at final inspection, rework it and release it. | The unit is held and cannot be picked, and the rework is recorded against it. | Held stock can be allocated to a shipment. |
| 10 | Dispatch 8 units. | The shipment is recorded by line and the order stays open for 2. | The whole order is marked shipped. |
| 11 | Invoice the dispatched quantity. | The invoice is for 8 units and references the shipment. | The invoice is for 10. |
| 12 | Close the work order. | Actual and standard cost are compared and variances post to the accounts finance designed. | A WIP balance remains with no explanation. |
| 13 | Trace one component lot forwards and backwards. | The trace finds the supplier receipt, work order, finished serials and customer shipment. | The trace needs a spreadsheet or a database query. |
| 14 | Reconcile at the end of the run. | On-hand stock, WIP and the ledger agree with the physical count and the expected postings. | A difference is noted to be explained later. |
How do you test migrated data before go-live?
Material planning works from item data, bills of materials, lead times, stock and open orders. If those are wrong, the plan is confidently wrong: it buys what is already on the shelf and misses what is short. Test the load as carefully as the configuration.
| Data | What to check | Reconcile against |
|---|---|---|
| Items and units | Every active item, unit of measure and conversion | The source item list, plus a physical check of a sample |
| Bills of materials and routings | Revision, quantities, scrap factors, operations and work centres | Engineering records, and a trial cost roll compared with known costs |
| Lead times and planning settings | Supplier lead times, lot sizes, safety stock and calendars | Buyer and planner sign-off for the main items |
| Opening stock | Quantity by item, location, lot or serial, and value | The physical count and the old system’s valuation |
| Open purchase orders | Supplier, item, outstanding quantity and date | Supplier confirmations for a sample |
| Open sales orders | Outstanding lines, prices and promised dates | The customer order book |
| Open work orders and WIP | Quantity complete, material issued and WIP value | A floor count and finance’s WIP balance |
| Balances | Customer and supplier open items and ledger opening balances | The trial balance at the cutover date |
Load the data at least twice in rehearsal. Compare record counts and totals, then spot-check individual records with the people who own them. Open work orders are usually the hardest item: decide early whether to migrate them part-complete, finish them in the old system, or close and reopen them, because each choice changes how WIP is valued on day one.
What are the go-live criteria for a manufacturing ERP?
Agree the go/no-go criteria before testing starts, so that the planned date does not decide the answer. A useful set for a manufacturer:
- Every core production-flow script has passed with migrated data, not demonstration data.
- No open defect stops the business shipping, invoicing or recording production. Each lower-severity defect has a written workaround and an owner.
- A timed mock cutover finished inside the planned window, and operations and finance signed its stock, WIP and ledger reconciliation.
- Each user role completed its acceptance scripts without help, on the devices and labels it will use.
- Failed integration messages land in a queue that a named person watches, and a retry does not duplicate an order or a posting.
- The fallback plan states the point after which the business will not return to the old system, and how it would ship and invoice if the new one stopped.
- Hypercare, the period of extra support after go-live, has named owners, a daily reconciliation and a daily defect review.
How should a manufacturer plan ERP testing?
- 01Define the flows and pass criteria
Pick the product families that cover your production models. Write the scripts and the go/no-go criteria before configuration ends.
- 02Build test data from real records
Use an extract of items, BOMs, routings, stock and open orders, so tests meet the same gaps and errors live use will.
- 03Test configuration and integrations
Include failures, duplicates and retries for every connected system, not only the successful message.
- 04Run the production flow end to end
Process owners run the script, record each result, and rerun it after every fix.
- 05Run acceptance on the floor
Real users, devices, labels and shift patterns, without a consultant narrating the steps.
- 06Rehearse the cutover twice
Time the data loads, reconcile stock, WIP and the ledger, and test the fallback.
- 07Decide, go live and watch
Hold the go/no-go meeting against the agreed criteria, then run hypercare and regression testing after every change.
Testing takes time from the same planners, supervisors and finance staff who are already running the factory, so book that time when the project plan is set, not when the date gets close. The aim is the same as in selection: a system that tells the truth about demand, material, work and cost. The ERP and manufacturing service describes how Harmatiq approaches that work, and the order-to-cash guide covers the handoffs once finished goods leave the factory.
Bring one real order, the systems it crosses and the exceptions your team handles manually. We can turn that journey into an implementation scope.