Master Data Management (MDM) is the practice of creating, maintaining, and governing a single, authoritative, "golden record" version of core business entities (customers, products, employees, locations, suppliers) across all systems in the enterprise. Without MDM, the same customer might be represented differently in Salesforce, the ERP, the web analytics platform, and the data lakehouse, making cross-system analysis unreliable.
Why MDM is Critical for Lakehouses
Data lakehouses ingest data from dozens of source systems simultaneously. Each source system has its own representation of core entities: a customer in Salesforce has a different ID format than the same customer in SAP, which differs again from the web analytics cookie ID. Without MDM resolution, analytical queries comparing customers across systems produce misleading results because the same individual is counted multiple times or not matched at all.
MDM Patterns
- Registry Style: A central registry maintains a cross-reference mapping of entity IDs across all source systems without copying the data. Each system keeps its own records; the registry provides the resolution keys for joining them.
- Consolidation Style: Master records are assembled from source system records using matching, survivorship, and merge rules. The golden record is published back to consuming systems and the lakehouse.
- Coexistence Style: Master records are published from MDM but source systems also maintain their own records. Useful when source systems cannot be modified to accept MDM IDs.
MDM and Iceberg
The MDM golden record dimension table (dim_customer_master, dim_product_master) is typically stored as an Iceberg table in the lakehouse's Gold layer. CDC pipelines keep it synchronized with the MDM system. All fact tables join to these master dimension tables using the MDM-resolved surrogate keys, ensuring consistent customer and product definitions across all analytical queries.
