Incentive programmes are unusually sensitive to master-data quality because they evaluate rules across aggregated volume. A customer duplicated across three records does not just look untidy — it splits volume three ways and causes the customer to miss a tier they earned.
The practical standard is not perfect master data, which nobody has. It is a cross-reference layer that can resolve imperfect data plus an exception queue for what it cannot.