Choose records by risk
| Sample | Why include it | Expected check |
|---|---|---|
| Active customer with leading-zero ID | Identity preservation | 00042 remains 00042 |
| Inactive customer with child records | Status and relationships | Status retained; children link correctly |
| Ambiguous source date | Meaning before conversion | Held for review until confirmed |
| Repeated external identity | Collision treatment | No silent merge |
| Multiline note in export | Export/import representation | Content stays within the intended field |
Set the pilot exit decision
In a hypothetical 30-row pilot, 28 accepted and two rejected can be an expected result if those two were deliberate invalid cases. Record why each rejection occurred. Require resolved critical mapping gaps, reconciliation of all inputs and an agreed repeat-run result before expanding the batch. A small successful sample is not a statistical guarantee about the entire dataset.
The pilot passes but the full export contains a new status code. What changes?
Show the worked answer
Hold that code for mapping review, add a representative case and repeat the affected checks. Keep the scope change visible rather than assuming it means the same as a known status.