
General Manager

The migration report shows 200 units in both systems. The warehouse totals match. The item totals match too. At the picking desk, the new system says one item is available in the wrong quantity at the wrong warehouse. All three summary checks can pass while the operational balance is wrong.
Before switching systems, I would ask the inventory owner to reconcile the level at which the team actually makes a decision. In this fictional example, that is item plus warehouse. An operation that picks by lot, bin or status would need to account for those distinctions in its own scope.
For this exercise, both extracts are meant to represent available units at 18:00 Riyadh. The agreed scope is two items and two warehouses. Quantity is in individual units. The source is the approved operating record at the cut-off. The target is the staged opening balance. The owner has confirmed the item and warehouse mappings and has agreed how to account for movements around the cut-off.
In a real migration, write down the source time and timezone, the target load or replication watermark, the units and status filters, and any permitted transformations. A source export from before the final movement and a target export taken after it represent different states. Reconcile that timing before assigning a difference to the migration.
| Warehouse and item | Source units | Target units | Difference |
|---|---|---|---|
| North, A | 60 | 55 | −5 |
| North, B | 40 | 45 | +5 |
| South, A | 30 | 35 | +5 |
| South, B | 70 | 65 | −5 |
| Total | 200 | 200 | 0 |
North and South each have 100 units in each extract. Item A totals 90 and item B totals 110 in both. Four item and warehouse combinations are still wrong. The signed differences cancel. The absolute differences add to 20 units across the four rows, which describes the sum of the discrepancies at this comparison level. The table alone does not establish how many physical units moved or what caused the mismatch.
Match records by the agreed key after applying approved mappings. Include keys present on only one side and report duplicates where the agreed balance key should be unique. If raw stock is legitimately split across several rows, aggregate to the agreed grain first and retain the detail for investigation. Keep a missing record separate from an explicit zero unless the business rule treats them as equivalent.
In this case, the acceptance rule is exact agreement on the four in-scope balances. Hold the inventory cutover. Ask the data lead to trace the four exceptions through the source extract, mapping and staged load. Ask the inventory owner to verify whether any movement near 18:00 explains a difference. Keep the original extracts and log the cause once it is confirmed. Correct through the approved process and repeat the comparison with the agreed boundary.
The acceptance record should show the cut-off, scope, key, row counts, missing or duplicate keys, the exceptions, their owner and the rerun result. Have the operational owner demonstrate a representative pick or reservation in the approved test environment as a separate check. A matching balance establishes the tested data condition. The operational rehearsal shows whether the team can use that state for the intended task.
AWS Database Migration Service documents a product-specific example of comparing corresponding source and target rows and reporting mismatches. Use your own tooling and data rules to support the business comparison. For an integration review, bring a representative balance, the cut-off and the decision that depends on it. Altaius Systems Integration.
Transform leadership capability with scalable, culturally intelligent simulation and measurable insights.