Post-load data validation techniques
Key Takeaway
Validate a data load with record-count reports, spot checks, and exception reports together, because each one answers a different question. Exception reports show which records were not migrated, so they tell you what to re-migrate.
Post-load data validation is the check you run after a data load. You use it to confirm that the expected records arrived and that the values look right, and to find which records did not load. In Agentforce Manufacturing, you can bulk load data such as products, assets, sales agreements, and forecast data from external systems. Validation tells you whether the load is done or whether you need to re-migrate or update data.
The three validation techniques
Salesforce recommends using all three of these techniques together to validate a migration, because each one answers a different question.
| Technique | Question it answers | What it shows |
|---|---|---|
| Custom reports on record counts | Did the expected volume arrive? | Record counts and an overall snapshot of the migration |
| Spot checks | Do individual records look right? | The values in the records you check |
| Exception reports | What failed to load? | Which data was not migrated |
You need the three techniques together to confirm that migrated data is complete and correct, because no one of them does that alone. A count report can match the expected total even when some values are wrong. A spot check sees only the records you choose to open.
A team migrates 5,000 asset records from a legacy system. The count report shows 4,980 assets, so 20 are missing. The exception report lists those 20 records, so the team knows which ones to re-migrate. Next, an admin named Priya opens ten assets and compares them with the legacy system. Three of them show a blank install date, but the legacy system has a date for each one. Those records did load, so the team updates them instead of re-migrating them. Without the spot check, the matching count after the re-migration would have hidden the blank dates.