Skip to content

Lesson 10 — Governance, Metrics, and Exam Decision Rules

Reference and Master Data are shared resources. Technology can store and move them, but governance determines the decision rights that make the shared environment trustworthy.

Governance decides

A Chapter 10 governance body should establish or oversee: - authoritative sources and source roles; - enterprise ownership and stewardship; - Data Quality rules and expectations; - conditions of use and sharing; - match thresholds and exception policies; - survivorship/trust rules; - Global ID authority; - Reference change approval gates; - monitoring frequency and evidence; - privacy, security, and retention requirements; - escalation and remediation; - enterprise adoption expectations.

The guiding principle is that shared Reference/Master Data belongs to the organization rather than one application or department.

Seven metric families

1. Data Quality and compliance

Answers: Does governed data meet business/rule expectations?

2. Data change activity

Answers: How often does data change, from where, and with what lineage? High activity can reveal volatility, workload, or matching-rule tuning opportunities.

3. Data ingestion and consumption

Answers: Which systems provide governed data, which consumers subscribe, and how much movement/use occurs?

4. SLA/service performance

Answers: Is shared data available and delivered within agreed service expectations?

5. Steward coverage

Answers: Which domains or datasets lack accountable stewardship?

6. Total Cost of Ownership (TCO)

Answers: What does the program/solution cost to operate and sustain?

7. Sharing volume / usage / adoption

Answers: Is the shared capability actually reused at enterprise scale?

Counts are not outcomes

“10 million records in the MDM hub” is inventory, not proof that the program is effective. A balanced view needs quality, service, stewardship, adoption, change, cost, and sharing evidence.

Seventeen high-value decision rules

  1. Reference vs Master: classifying/allowed values vs persistent entity identity.
  2. Master vs Transaction: context/entity involved in activity vs event itself.
  3. Reference/Master vs Metadata: business values/entities vs information describing/managing the data.
  4. List vs Cross-reference vs Taxonomy vs Ontology: allowed set → translation → hierarchy → richer semantic relationships.
  5. SOR vs System of Reference: authoritative create/maintain vs authoritative consume/access.
  6. Trusted Source vs Golden Record: governed best source/view vs one reconciled entity record.
  7. Standardization vs Entity Resolution: normalize evidence vs decide same/different entity.
  8. False positive vs false negative: different entities joined vs same entity left separate.
  9. Deterministic vs probabilistic: explicit repeatable rules vs statistical likelihood.
  10. Duplicate identification vs match-link vs match-merge: flag → link identity → reconcile content.
  11. Global ID vs Source ID vs X-Ref: enterprise identity vs local identity vs mapping/history.
  12. Affiliation vs parent-child: flexible typed association vs fixed hierarchy.
  13. Registry vs Transaction Hub vs Consolidated: point → own writes → copy/reconcile.
  14. Enterprise MDM model vs source schema: shared semantics vs local “system speak.”
  15. MDM tool vs MDM discipline: installed capability vs people/process/governance/technology operating model.
  16. Local Reference edit vs governed enterprise change: isolated adjustment vs shared-impact controlled workflow.
  17. Repository population vs program success: stored records vs trusted, consumed, governed outcomes.

Common distractor traps

  • “Golden” = perfect truth. False.
  • “Reference Data needs entity resolution.” False.
  • “Registry centralizes writes.” False.
  • “Consolidated is instantly current.” Not necessarily; replication creates latency.
  • “Match-link changes source attributes.” False; match-merge reconciles content.
  • “One department owns shared Master Data.” Enterprise governance is required.
  • “A small Reference change can be made locally without impact review.” Shared impact may make that unsafe.

Final mental model

Govern shared classifications through RDM. Reconcile persistent entity identity through MDM. Define authority and architecture. Preserve quality, identifiers, mappings, history, and reversibility. Distribute trusted data through stewardship and sharing. Measure whether the enterprise actually uses and benefits from it.

Source anchor: pp. 329–359.

← Lesson 9 · Exam Map →