How to check business data after a system migration
A success message is only the start of verification
After data moves from an old application to a new system, the screen may say that import is complete. Users still need to establish whether records are complete, accurate and usable. Agree on acceptance checks before migration and involve data owners. A working application or the absence of error messages should not be the entire acceptance criterion.
Define the source dataset and cut-off time
Specify the data types, date range and whether cancelled or unfinished records are included. Retain the reference source dataset and its export time. If people continue adding or changing records in the old system, plan how later changes will be handled. Separate differences caused by different time windows from migration errors, rather than adjusting totals to match without finding the cause.
Equal counts do not establish matching records
In a hypothetical example, the source contains 100 customers and the destination also contains 100, but one customer is missing and another appears twice. The count still matches. Compare agreed identifiers and check for missing or duplicate records. Microsoft’s preview migration-tool documentation distinguishes count reconciliation from field-level checks. This article uses that distinction to explain the principle, not to recommend a preview feature for production.
Check values and relationships
Select samples covering normal and exceptional records, such as Thai names, blank dates, negative amounts or customers with several addresses. Check that documents remain linked to the correct customer and that attachments within the migration scope can be opened. Sampling can reveal problems but does not prove that every record is correct. Define stronger checks for important data according to risk.
Track outstanding differences and involve the owner
List differences with reference identifiers, impact, assigned owners and retest methods. Clearly separate issues that must be fixed before launch from those the business owner explicitly accepts for later follow-up. Do not remove errors from the report to make it look complete. Do not rerun an import without knowing whether it creates duplicates or updates existing records. Retain results from each run for traceability.
Accept the work using evidence of usability
Before closing the task, have users search for records, open documents and perform sample tasks in their actual roles. Summarise completed checks, remaining limitations and responsibilities after launch. To discuss system migration or integration with Chawinnix, prepare system names, data types and desired acceptance criteria. There is no need to send a real customer database during an initial enquiry.
