Your inventory lead signs off the weekly promotional file from the marketing team without checking the SKU master first. The upload creates new records for every variant instead of updating existing ones because the matching keys use internal product codes that marketing never sees.
The ERP reconciliation job runs at 2 a.m. and flags the overlaps, yet the overnight SLA forces the team to accept the new entries rather than delay the price changes. Those extra SKUs now carry forward into the planning system as separate demand streams.
Three weeks later the forecasting engine shows a sudden spike in a core product line that never happened. The model splits historical volume across the original SKU and the new duplicate, so safety stock calculations swing wildly for both.
Your replenishment planners start overriding suggested orders because the numbers look wrong. They burn hours manually merging the records every Monday while the vendor charges extra for each “custom deduplication rule” that still fails on the next promo batch.
The single source of truth claim on the dashboard stays green because it only measures records inside the MDM tool, not the downstream systems that actually drive purchasing. The AI never sees the merged reality.
Fix the intake instead. Force every promotional file through the same matching service used for new supplier items, and reject any upload that would create an unapproved duplicate. Month-end bonuses for marketing then tie to clean master data rather than campaign launch speed.