Worked example. Demonstrated on sample material master data. Ready to apply to your system.
The problem
Duplicate materials tie up cash in stock, split purchase history and send planners to the wrong part. Most duplicate tools are black boxes, so planners do not trust the results and the clean-up stalls.
What we built
A rule-based, explainable process that goes from a system extract to a merge plan your team can execute. Fast, certain rules run first, and AI is used only for the hard cases.
- Collect: material, vendor, equipment, BOM and posting data is extracted. Nothing is changed in the system.
- Normalise: part numbers, descriptions, units and manufacturer names are cleaned so like is compared with like.
- Match: seven plain rules find duplicates and near-duplicates, each with a stated confidence.
- Explain: every finding names the rule, the records and the evidence.
- Plan: for each group, which record survives, what is re-pointed and what is retired.
- Review and fix: your team confirms each group. Duplicates are blocked and usage migrated, never deleted while anything still points at them.
What a finding looks like
Two records with part numbers FLT-TRN and 00FLT-TRN, the same manufacturer, material group and unit, and an identical description, copied typo included. Confidence: high. A planner can check it in ten seconds. Look-alikes, such as sizes that differ by one digit or left- and right-hand parts, are deliberately left alone.
Related service: Master Data Remediation · Talk to us
