Use it before the first load
Required field validation belongs near the start of the ETL workflow. If an order number, customer code, invoice date, or account ID is empty, there is no point loading the row and hoping a report will forgive you later.
The validation rule stops the record, logs the issue, and sends the bad row to an exception path. That keeps the original data untouched and gives support staff something useful to inspect.
What to validate
- Primary business keys such as customer ID, invoice number, or order number.
- Dates needed for reporting, scheduling, or reconciliation.
- Status fields used by later workflow decisions.
- Any field used in joins, lookups, or database constraints.
When not to use it
Do not mark every field as required just because the form looks tidier that way. Optional notes, secondary phone numbers, and unused legacy columns remain optional. Over-validating data creates noise, and noisy validation logs are where good rules go to retire.
Useful references
For broader context, see the three types of data validation, chaining multiple data validations, and IBM's overview of data validation.