Deep Understanding, Selective Memorization & Traps
Must understand deeply
- Reference Data is generally smaller/less volatile than Master Data but still changes and requires governance. Entity resolution is a Master Data identity problem, not an RDM problem.
- Master Data is not “all important data.” Transactions and Reference domains can be critical without becoming Master Data.
- Golden Record is not perfect truth. Use Trusted Source language when you need to emphasize governed best-available data without overclaiming certainty.
- Enterprise Master modeling must overcome source semantics. A canonical/logical enterprise entity model should not simply inherit the vocabulary/granularity of one application.
- Match-merge is harder to reverse than match-link because it changes the effective master content through survivorship decisions.
- Match history is a governance requirement. It supports auditability, unmerge/remerge, metrics, exception analysis, and rule improvement.
- Global IDs and X-Ref history solve different problems. The ID is the enterprise identity; X-Ref is the traceable mapping to local IDs.
- Affiliations can be more flexible than parent-child trees. Use typed relationships when many-to-many or multiple contextual views matter.
- Architecture choice follows authority and consumer needs. Consider SOR count, organizational structure, governance maturity, latency, source disruption, and consumption patterns.
- Central RDM creates value only when reused. A populated repository with continued local code sets is an adoption/governance failure.
Selective memorization
Reference structures
List → Cross-reference → Taxonomy → Ontology.
Reference source/type families
Internal/proprietary; industry/standards; geographic/geo-statistical; computational/other governed domains.
Common Master domains
Party; Product; Location; Financial; Legal; enterprise-specific directories/entities.
Tool families
Integration/remediation and DQ tools; operational/shared stores; data-sharing hubs; workflow/stewardship; specialized MDM applications.
Metric families
DQ/compliance; change activity; ingestion/consumption; SLA performance; steward coverage; TCO; sharing/usage.
Distractor traps
“MDM = an MDM software product.”
Wrong. MDM is a people/process/governance/technology discipline.
“Golden means 100% accurate and complete.”
Wrong. The reconciled record can still contain unknown or imperfect attributes.
“Reference Data requires entity resolution.”
Wrong. RDM controls domains, values, mappings, definitions, hierarchy, and change.
“Registry centralizes Master Data writes.”
Wrong. Registry indexes source-held data; Transaction Hub centralizes writes.
“Consolidated hub is instantly current.”
Wrong. Replication introduces latency/synchronization considerations.
“Match-link changes source values.”
Wrong. Match-link establishes identity relationships; match-merge reconciles attributes.
“Shared Master Data can belong to one department.”
Weak. Shared data needs enterprise governance and defined stewardship.
“Small Reference changes can be made locally.”
Dangerous. A small code/hierarchy change can affect many downstream consumers.