Master Data Management for Manufacturers: A Practical Guide
Master data management for manufacturers is the discipline of maintaining one trusted definition of item, customer, supplier, site, and BOM data across every system that consumes it. It matters most in companies that grew by acquisition and now run two or three ERPs, because the same physical part exists under different numbers in SyteLine, Infor LN, and a legacy Baan instance. Without MDM, consolidated spend analysis is guesswork, cross-plant sourcing is impossible, and every merger integration restarts the same reconciliation from scratch.
Which Domains to Tackle First
Sequence by economic value, not by data volume. Supplier and item are almost always first for discrete manufacturers because they unlock spend consolidation and cross-plant sourcing. A mid-size company with three plants typically discovers that 20 to 35 percent of purchased part numbers exist in more than one ERP under different identifiers, and that the same supplier appears as four vendor records with different payment terms. Customer master comes next for companies with complex hierarchies, where the same aerospace prime is a dozen ship-to records with no parent link, making true customer profitability invisible. Site, plant, and cost center follow because they are small, stable, and unblock consolidated financial reporting.
Registry, Consolidation, or Centralized Style
Three implementation styles exist and the choice determines cost. A registry style leaves records in each ERP and maintains cross-reference keys plus a golden view for analytics; it is the fastest to deliver, often three to six months, and does not disturb transactional processes. A consolidation style physically builds golden records in a hub for reporting only. A centralized style authors master data in the hub and publishes it to each ERP, which delivers the cleanest long-term outcome but requires changing how every plant creates parts and usually runs twelve to twenty-four months. Most manufacturers should start with registry for analytics, then move item creation to a centralized workflow for new parts only.
Matching, Survivorship, and Cross-Reference Keys
The technical core of MDM is deciding which record wins. Matching combines deterministic keys such as manufacturer part number or DUNS number with probabilistic scoring on descriptions and attributes. Survivorship rules then decide, field by field, which source supplies the golden value: engineering attributes from PLM, commercial terms from the ERP that owns the contract, and classification codes from a controlled reference list. Persist the cross-reference so every analytical fact can roll up to the golden entity without rewriting source systems. Tune the match threshold deliberately: in item master work, a false merge is far more damaging than a missed one, because merged parts corrupt BOMs and open orders. Aim for high precision, send the ambiguous middle band to a steward queue, and measure both false positives and unmatched volume every month.
- Deterministic first: exact match on manufacturer part number, CAGE code, DUNS, or GTIN where available
- Probabilistic second: weighted similarity on description, specification text, and commodity classification
- Field-level survivorship: PLM wins on engineering attributes, ERP of record wins on commercial terms
- Persist a cross-reference table mapping every source key to the golden ID, with effective dates
Governance, Workflow, and Defense-Specific Attributes
MDM lives or dies on the create and change workflow. Define a new-part request with required attributes by commodity type, an approval path through engineering and planning, and an SLA measured in hours rather than days, because a slow process pushes engineers back into local part creation. Defense and aerospace suppliers must govern additional controlled attributes such as ECCN and USML category, country of origin, DFARS specialty metals compliance, and export-control flags. These attributes must be stamped on the golden record, since a misclassified part is a compliance event, not a data issue. Give each domain a steward with real authority to reject a request, and publish cycle-time metrics for the workflow itself so slow approvals are visible. Governance that only exists as a policy document, without measured SLAs and a queue someone owns, reverts to local part creation within a quarter.
- New-item request workflow with commodity-specific required attributes and a 24-hour approval SLA
- Controlled attributes governed centrally: ECCN, USML category, country of origin, and DFARS flags
- Change requests versioned with effective dates so historical reporting stays reproducible
- Quarterly stewardship review of match rules, false positives, and unmatched record backlog
How Netray Accelerates Manufacturing MDM
Netray uses AI agents to do the work that traditionally consumes the first year of an MDM program. Agents parse item descriptions, specification text, and drawing metadata across SyteLine, LN, M3, and legacy Baan, cluster likely duplicates, and propose golden records with per-field source attribution and a confidence score. Stewards approve or reject in a review queue instead of building spreadsheets. Agents also classify parts to UNSPSC or an internal commodity taxonomy and flag missing export-control attributes. A typical engagement reaches a usable cross-plant item cross-reference in eight to twelve weeks, and for ITAR-constrained clients runs fully on-prem with local models.
Frequently Asked Questions
Do we need an MDM tool or can we do this in the ERP?
If you run a single ERP instance with one plant, disciplined governance inside the ERP plus data quality rules is usually enough. The moment you have two or more ERPs, an acquisition pipeline, or PLM as a separate authoring system, you need a hub outside the ERP to hold cross-references and golden records. That hub can start as a modest set of tables in your warehouse rather than a large commercial platform.
How long does a manufacturing MDM implementation take?
A registry-style implementation for one domain, typically supplier or item, delivers usable cross-references and a golden view in three to six months. Moving authoring into a centralized workflow that publishes back to each ERP usually takes twelve to twenty-four months because it changes how every plant creates data. Sequencing by domain and delivering analytical value first keeps sponsorship alive through the longer phases.
What is survivorship in master data management?
Survivorship is the rule set that decides which source system supplies each field of a golden record when several systems disagree. It is applied per field, not per record: engineering attributes may survive from PLM, payment terms from the ERP holding the contract, and classification codes from a governed reference list. Good survivorship rules also record which source won, so any golden value can be traced back to its origin.
Key Takeaways
- 1Which Domains to Tackle First: Sequence by economic value, not by data volume. Supplier and item are almost always first for discrete manufacturers because they unlock spend consolidation and cross-plant sourcing.
- 2Registry, Consolidation, or Centralized Style: Three implementation styles exist and the choice determines cost. A registry style leaves records in each ERP and maintains cross-reference keys plus a golden view for analytics; it is the fastest to deliver, often three to six months, and does not disturb transactional processes.
- 3Matching, Survivorship, and Cross-Reference Keys: The technical core of MDM is deciding which record wins. Matching combines deterministic keys such as manufacturer part number or DUNS number with probabilistic scoring on descriptions and attributes.
Put this into numbers
Free interactive tools for exactly this problem. No signup to use them.
AI Governance Maturity Assessment
Score your AI governance across policy, inventory, risk classification, data handling, monitoring, and executive oversight, and get a banded improvement roadmap.
Free ToolPoor Data Quality Cost Calculator
Put a defensible annual dollar figure on bad data using error volume, rework time, downstream escalation multipliers, and revenue leakage.
Free ToolAI Training Data Readiness Assessment
Score your data across volume, machine readability, labeling, lineage, permissions, rights, and refresh so you know what to fix before building an AI system.
Terms used in this article
Contact Netray to run an AI-assisted item and supplier match across your ERP instances and see your true duplicate rate in weeks.
Related Resources
An ERP Data Quality Framework
An ERP data quality framework for manufacturers: profiling item and BOM data, six quality dimensions, automated rules, scorecards, and remediation workflows.
AI & AutomationERP Data Warehouse Architecture
ERP data warehouse architecture: landing, staging, and star schema layers, CDC extraction from SyteLine and Infor LN, plus governance for manufacturers.
AI & AutomationNatural Language Query Over ERP Data
Natural language query over ERP data: how text-to-SQL, semantic layers, and RAG let SyteLine and Infor LN users ask questions and get trustworthy answers.