Rapid Recall Key 01–33
- Reference and Master Data Management: discipline of reconciling, governing, maintaining, and sharing critical Reference/Master Data so it is consistent, current, authoritative, and reusable.
- Goals: complete/consistent/current/authoritative shared data; enterprise reuse/sharing; lower risk/cost from inconsistent values, identities, and definitions.
- Guiding principles: enterprise ownership, authority, quality, stewardship, controlled change, accountability/governance for shared use.
- Four-way: Reference = controlled classification/domain values; Master = persistent entities; Transaction = events; Metadata = source/definition/steward/lineage/lifecycle information about data.
- Reference Data is generally a smaller controlled domain and changes less often than entity populations/attributes, although it still changes under governance.
- RDM governs controlled values/mappings; entity resolution asks whether records represent the same real-world entity, which is MDM.
- Reference Data: controlled data used to characterize/classify other data or relate internal information to standards/external information.
- Structures: List, Cross-reference, Taxonomy, Ontology.
- List: one controlled value set. Cross-reference: maps alternative representations/code sets of the same concepts.
- Cross-reference maps equivalents across sets; taxonomy organizes concepts hierarchically.
- Taxonomy expresses classification/hierarchy; ontology represents richer formal concepts and relationships.
- Examples: internal/proprietary, industry/standards, geographic/administrative, computational/other governed domain sets.
- Definition, steward/owner, source authority, version/effective dates, update schedule, consumers, history, mappings, status/deprecation.
- Master Data: persistent data identifying/describing core business entities reused across processes and systems.
- Party/customer/supplier, Product/service, Location, Financial/account structures, Legal entities, and enterprise-specific core domains.
- SOR: authoritative create/maintain. System of Reference: authoritative consumer access.
- Trusted Source: governed best-available source/view. Golden Record: reconciled record for one entity instance.
- “Golden” can imply perfect truth, but a reconciled record can still contain incomplete/unknown/imperfect attributes.
- MDM needs people, process, governance, stewardship, DQ, architecture, integration, and technology; installing software alone is not MDM.
- Ask: what entities/attributes matter; where created; where stored; how changed; who uses/accesses; what quality/reliability issues exist; what authority/sharing/integration expectations apply.
- Source landscapes, semantics, identifiers, DQ, and governance are complex; manageable domain scope makes reconciliation/stewardship/value delivery controllable.
- Model → Acquire → Validate/Standardize/Enrich → Resolve Identity + Manage IDs → Share + Steward.
- Data Model Management defines enterprise entity/attribute semantics and granularity beyond source-system “system speak.”
- Acquisition brings source records into MDM; standardization normalizes their representation for comparison/use.
- Standardization normalizes what exists; enrichment adds trusted supplementary information.
- Entity resolution determines whether multiple records/references represent the same real-world entity or different entities.
- False positive = different entities incorrectly joined; false negative = same entity incorrectly left separate.
- Similarity analysis compares candidate records using relevant attributes/rules/scores to estimate same-entity evidence.
- Deterministic = explicit fixed rules/repeatable outcome; probabilistic = statistical likelihood based on weighted/trained evidence.
- Candidate identification finds records worth comparing; identity resolution makes the final same/different decision.
- Match history supports auditability, metrics, rule improvement, and reversal/unmerge/remerge when a decision proves wrong.
- Duplicate identification flags possible duplicates without identity change; match-link establishes identity/X-Ref while preserving source attributes.
- Match-link connects identity; match-merge reconciles attributes into a unified record.
Source: Chapter 10, pp. 329–359.