Most data-quality programmes are too reactive.
Issues are often discovered:
after a report is published
after a user complains
during reconciliation
during migration
during audit
when an AI output exposes bad source data
By then, the cost is already higher.