Master Data Management: Customer, Product, and Reference

Master data management (MDM) coordinates shared data about important entities and classifications across systems. Customer data integration (CDI), identity resolution (IR), product information management (PIM), and reference-data management address related but distinct domains.

Domain distinctions

Domain	Typical entities	Key risks
Customer/party	People, households, organizations, accounts, relationships	False matches, privacy, consent, protected traits, temporal identity
Product	Products, SKUs, offers, bundles, classifications	Variant confusion, inaccurate claims, channel inconsistency, supplier provenance
Reference data	Codes, taxonomies, countries, currencies, statuses	Version drift, invalid mapping, effective-date and jurisdiction errors
Location/supplier/asset	Sites, vendors, equipment, legal entities	Hierarchy, ownership, sanctions, safety, duplicate identifiers

Choose an architecture pattern

Registry, consolidation, coexistence, and centralized patterns describe where mastered attributes live and which systems author them. Choose from latency, ownership, migration, availability, regulation, and consumer needs. No pattern automatically creates a single source of truth.

Match entities with measurable uncertainty

Standardize relevant fields, generate candidates, compare attributes, assign confidence, and route ambiguous cases for review. Measure false merges and false splits, precision/recall, review workload, subgroup behavior, and reversibility. Identity resolution is inference, not proof that records describe the same person.

Define survivorship and provenance

For each mastered attribute, define source authority, validation, recency, completeness, effective dates, confidence, exceptions, and stewardship. Preserve source values and transformation lineage. “Most recent wins” is not safe when timestamps, synchronization, or sources are unreliable.

Govern hierarchies and reference data

Model relationship type, direction, validity period, cardinality, and jurisdiction. Prevent invalid cycles and orphans, version code sets, publish effective dates, and provide mappings for consumers. Apply data-access governance at attribute, record, and workflow levels.

Integrate and reconcile

Specify identifiers, schemas, contracts, ordering, idempotency, retries, dead-letter handling, freshness, version compatibility, and deletion. Reconcile publishers and consumers; a successful API response does not prove downstream consistency. See data-integration challenges.

Measure operational outcomes

• Coverage of governed entities and attributes.
• Duplicate candidates, confirmed false merges/splits, and reversal time.
• Quality-rule failures with denominators, severity, and exception age.
• Hierarchy/reference version conflicts and consumer reconciliation.
• Steward workload, decision time, and recurring root causes.
• Access exceptions, privacy requests, incidents, and recovery tests.

For applied scenarios, use the MDM interview questions. MDM succeeds when shared decisions and controls improve measurable outcomes—not when every value is centralized.

Originally published February 25, 2021; technically reviewed and substantially updated September 4, 2026.