People need to trust that the work waiting in a new system is the work they were already responsible for. That trust is a deliverable of migration, alongside the application and the transferred records.
Consider a hypothetical move from an aging customer database into a new portal. The migration log reports success. The row counts agree. On Monday, an account manager opens the portal and discovers that a customer with several locations appears to have only one active delivery address.
All the rows may exist. The business relationship between them has been damaged. Counting boxes is a limited way to establish whether anyone packed the contents correctly.
Define what agreement means
A reconciliation contract is a practical acceptance agreement: what must match, which transformations are intentional, what exceptions are acceptable, and who can accept them. It can live in the existing migration plan; the name does not require another governance layer.
Begin with the critical journeys. Can staff find the correct customer? Can they distinguish active commitments from historical ones? Do balances and statuses retain their intended meaning? Can authorized users access the records they need?
Then turn those questions into checks. Record counts help detect missing records. Field comparisons can detect changed values. Business totals and relationship checks test whether those records still describe the same operating reality.
Use technical validation for the job it covers
AWS Database Migration Service validation provides a concrete example: it compares source and target rows and reports mismatches, with support and limitations depending on the endpoints and task. Validation also consumes resources and takes time.
That is useful evidence about transferred data. It does not establish that a new workflow interprets that data correctly, that every permission is appropriate, or that the business has accepted a changed status definition.
Plan for both technical comparison and business acceptance. Budgeting only for moving the data leaves the proof of correctness unfunded.
Make transformations explainable
Perhaps the old system stores several versions of a customer name. Perhaps the destination introduces a new identifier. Perhaps a retired status becomes a historical attribute.
Those changes may be sensible. Document the mapping and preserve enough lineage to trace a destination record back to its source. An unexpected difference should be distinguishable from an approved transformation.
For the portal example, the acceptance evidence should show how customer, location and delivery relationships survived. A sample of ordinary customers alone is weak evidence if the troublesome cases involve merged accounts, old identifiers or incomplete addresses.
Select examples that exercise the real complexity. Record what they cover and what they leave untested.
Decide where new work belongs
Incremental migration introduces a further question: which system is authoritative for each operation while the old and new systems coexist?
Martin Fowler's Strangler Fig discussion describes gradual replacement and the transitional architecture needed to support coexistence. The operational implication is worth making explicit: a team needs a clear place to create, change and resolve each kind of record during the transition.
Otherwise a successful rehearsal can be overtaken by new activity before cutover. Specify how changes after the validation checkpoint are captured, how the final comparison works, and what happens if the window closes with unresolved differences.
Write the acceptance receipt before the move
For the portal example, a small acceptance record can make the agreement testable:
- Population and checkpoint: active customer accounts, their locations and open delivery commitments at the named cutoff.
- Relationship check: each delivery remains attached to its intended customer and location, including merged accounts and retired identifiers.
- Approved transformation: record the source-to-destination identifier mapping and the person who accepted it.
- Exception decision: identify each unresolved mismatch, its operational consequence, its owner and the decision to fix, defer or stop cutover.
- Final verification: reconcile changes since the checkpoint and confirm that authorized staff can complete the critical journey.
These are illustrative acceptance fields, not a universal test suite. Choose the populations and checks from the actual operation. A balance check, for example, also needs agreed currency, rounding and timing rules before a difference can be classified as an error.
Give acceptance a visible finish
The final receipt should identify the compared populations, the checkpoint, the remaining exceptions and the authorized acceptance decision. Preserve evidence rather than flattening every outcome into “migration complete.”
Our broader legacy database migration guide covers the migration decision and delivery approach. This acceptance record provides a narrower way to prove that the destination preserves the agreed business meaning. For a migration, extend that ownership through the final reconciliation and into the first operating period.
If you are planning application modernization, bring the business records your team must trust on day one. They provide a concrete starting point for scope, validation and cutover decisions.
The account manager opening the portal on Monday should inherit a system they can explain, with the evidence behind it. That is a better finish line than a transfer progress bar reaching the end.





