Data Mistakes

Master Data Governance Mistakes That Corrupt Your ERP

The costly and recurring master data errors that quietly corrupt manufacturing ERPs, with the symptom, root cause, and a numeric fix for each.

The most expensive master data mistake is the silent unit of measure error. Symptom: a purchase order for 5 rolls of tape arrives as 5,000 meters, or MRP demands 12,000 fasteners when the BOM meant 12 boxes. Root cause is a UOM conversion factor entered as 1:1 when the real base-to-purchase ratio is 1:250, or a decimal shifted three places. The fix: run every item with more than one UOM through a Unit of Measure Conversion Risk check, prioritize the top 3 percent by transaction volume, and validate that base UOM times conversion equals the alternate UOM on 20 sampled receipts before you trust the field.

Duplicate items are the second wound, and they bleed slowly. Symptom: two part numbers for the same M6x20 bolt, one showing 4,000 on hand and the other showing zero and a rush PO. Root cause is no matching rule on manufacturer part number plus a free-text description that lets a steward type SCREW, screw, and Screw M6. Teams routinely carry 8 to 15 percent duplicates in an item master over 50,000 records. Run a Duplicate Item Cost pass, fuzzy-match on normalized description and MPN, and merge. Each duplicate typically ties up 200 to 900 dollars in phantom safety stock and split demand.

BOM and routing that disagree is the classic mistake that survives audits because each object looks fine alone. Symptom: the BOM consumes a component at operation 30, but the routing has no operation 30, so backflush fails or posts to the wrong work center. Root cause is engineering revving the BOM without revving the routing, or an operation deleted on one side only. Check every active production part through the BOM Routing Mismatch tool: if operation references on the BOM do not have a 1:1 match in the routing, flag it. A shop with 2,000 active items commonly finds 60 to 140 mismatches on first scan.

Stale work center data quietly wrecks capacity math. Symptom: finite scheduling shows a cell at 140 percent load while it sits idle, or lead times inflate by 3 to 5 days for no reason. Root cause is a work center still carrying a queue time of 480 minutes from a line that was rebalanced two years ago, or an efficiency of 100 percent on a cell that actually runs at 82 percent. Score every center with Work Center Data Completeness and flag any missing rate, efficiency, or calendar. Fix the fields feeding scheduling first: setup, run rate per hour, and available capacity.

Assuming the item master is clean because ERP go-live was clean is a process failure, not a data failure. Symptom: data quality was 95 percent at implementation and nobody has measured since, yet planners keep overriding MRP by hand. Root cause is entropy: 1 to 2 percent of records degrade per month through part changes, supplier swaps, and half-finished edits, so a two-year-old master can be 30 percent stale. Re-run an ERP Data Quality Score quarterly and an Item Master Cleanup Effort estimate before any migration. Treat any completeness score under 90 percent as an active liability, not a backlog item.

Wrong classification is the mistake that hides all the others. Symptom: spend analysis cannot roll up because 4,000 items sit in a MISC or UNCLASSIFIED bucket, and ABC analysis puts a 12-cent washer in class A. Root cause is optional classification fields, no controlled value list, and no owner. When more than 5 percent of items lack a valid commodity or ABC code, sourcing and planning both make decisions on noise. Size the backlog with the Item Classification Workload calculator, enforce a mandatory pick-list, and reclassify high-value items first. Getting the top 20 percent of spend correctly coded recovers most of the analytical value.

Understaffing stewardship guarantees the queue never drains. Symptom: new-item requests take 6 to 9 days against a 24-hour target, and engineers create shadow spreadsheets to work around the delay. Root cause is treating governance as a part-time duty: one steward per 40,000 to 60,000 active items is a realistic ratio, and many shops run at one per 150,000. Use Data Stewardship Capacity and Master Data Governance Workload together to compare inbound request volume against real throughput. If arrival rate exceeds service rate by even 10 percent, the backlog grows without bound; fix staffing or automate validation before adding rules.

Skipping routing accuracy verification lets bad standards compound into bad costs and bad schedules. Symptom: standard cost variances run 15 to 25 percent, and the same job is always late even after expediting. Root cause is routings with copied-forward times that were never re-studied: a 45-minute operation that now takes 28, or a scrapped step still consuming labor. Score routings with the Routing Accuracy Score tool, sample 10 high-runner parts, and time-study them against the standard. If measured time differs from the routing by more than 10 percent on a third of samples, freeze cost rollups until the routings are corrected.

Published 2026-07-01.