Data migration test cycles: university library with tall windows and reading tables

Test cycles, wipe and reload, and cutover

Why every data migration test cycle begins by clearing and reloading the migrated records, how the cycles are sequenced, and what happens over the go-live weekend.

Loading data into a test environment once and testing it repeatedly gives false confidence, because each round of fixes lands in an environment that no longer resembles production. Instead, migrated records are cleared from the Salesforce sandbox and reloaded from source before each test cycle, so every cycle runs the complete, current scripts against current data and starts from a known state. Clearing migrated records does not undo changes to records that were already in the org, such as merged duplicates; to repeat deduplication from scratch, the sandbox is refreshed from production, and because Salesforce limits how often a sandbox can be refreshed, each refresh is planned into the schedule.

What changes between data migration test cycles

Between cycles the mapping is refined, scripts are corrected and the exception report is worked through; the objective is a final cycle in which the exception report contains only records the client has agreed to exclude.

Diagram of the Salesforce load cycle repeated in each data migration test cycle
The load cycle, repeated in every data migration test cycle and again for production.

A typical path from kickoff to go-live

  1. The client signs off requirements and retention criteria, and approves the preliminary mappings.
  2. Source data is staged into the migration database and profiled: integrity, missing primary keys, duplicates.
  3. Scripts are developed iteratively per object and run against the test environment.
  4. Each cycle starts from a cleared and reloaded sandbox, followed by data validation and, where features are affected, functional and regression testing.
  5. Exceptions are dispositioned; mappings and scripts are corrected; the cycle repeats until accepted.
  6. End-to-end user acceptance testing follows, in which the client's users run complete business processes on migrated data; like the test cycles, it has no fixed number of rounds.
  7. A cutover runbook is written: sequence, timing, freeze windows, rollback conditions, delta strategy, go or no-go criteria. Before go-live, the sequence is rehearsed and timed, and the sign-off sheet is reviewed with the client.
  8. Over the go-live weekend, access to source systems is paused, data is pulled and migrated to production in the validated sequence, and delta loads are run where appropriate.

Two disciplines protect production. The first is that production repeats exactly the sequence validated in the final test cycle; nothing is done in production that was not done in test, except the production-only steps written into the cutover runbook. The second is a stabilization period after go-live, and we advise clients to keep the migration environment until it ends.

Assess the data before the agents depend upon it

The first call is with the practitioners who would do the work, not a sales intermediary. The measures come first and the estimate follows.

Talk to our experts