Reference and Master Data Management
Reference and Master Data Management
This domain tests the data that other systems depend on: shared business entities and controlled value lists. Questions focus on the difference between reference and master data, matching and survivorship, MDM implementation styles, and the governance that makes a golden record credible.
Core rules and concepts
- Reference data is a controlled set of permissible values or classifications used by multiple processes: country codes, currency codes, status codes, product taxonomies. It is often small, stable, and externally standardized, for example ISO 3166 country codes and ISO 4217 currency codes.
- Master data describes core business entities: parties such as customers, suppliers, and employees; products; locations; and financial structures such as cost centres. It is shared, relatively stable, and non-transactional.
- MDM creates and maintains a trusted, consistent view of master data across systems. A golden record is the trusted version, assembled by rules, not necessarily a single physical row in one source system.
- Entity resolution: standardize and cleanse, block candidate pairs, match using deterministic rules such as tax ID or email, or probabilistic scoring on name and address, apply thresholds, and review false positives and negatives.
- Survivorship rules decide which source value wins per attribute, for example CRM for contact details and ERP for billing details. Rules must be documented, versioned, and governed.
- MDM styles: registry (index and cross-reference only), consolidation (analytical hub), coexistence (bidirectional with sources), and centralized or transactional hub, where MDM is the system of record. Choice depends on latency, ownership, and operational risk.
- Reference data management needs ownership, change workflow, versioning, cross-reference mapping, and distribution to subscribers.
Worked example
CRM holds Jon Smith, 12 High St and ERP holds John Smith, 12 High Street. A deterministic match on tax ID finds no hit because CRM lacks it. Probabilistic matching on normalized name and address scores 0.91, above the 0.85 auto-match threshold, so the records are linked with a cross-reference. Survivorship gives CRM the email and phone, ERP the billing address and credit terms, and the golden record carries both source IDs. The steward reviews matches between 0.70 and 0.85. Publishing the golden record to the data warehouse and to customer service keeps them aligned, while the governance council approves the survivorship policy.
Traps that cost marks
- Confusing reference data with master data; code lists are not customer entities.
- Assuming the golden record must overwrite source systems; registry and consolidation styles do not.
- Setting a matching threshold without measuring false positives and false negatives.
- Letting IT choose survivorship rules without business owners; rules encode business priority.
- Ignoring hierarchies and relationships, which are part of master data, not an afterthought.