Agile, auditable enterprise data warehouse methodology
Data Vault 2.0 is a hybrid approach to modeling enterprise data warehouses that combines the best aspects of third normal form (3NF) and dimensional modeling. Developed by Dan Linstedt in the 1990s and refined into version 2.0, it emphasizes scalability, flexibility, and auditability while supporting agile development practices.
Data Vault is designed specifically for enterprise data warehouses that need to integrate data from multiple sources while maintaining complete historical tracking and adapting quickly to changing business requirements. It separates business keys, descriptive data, and relationships into distinct table types.
Core Philosophy:
Data Vault prioritizes adaptability and auditability over query performance. It is optimized for data warehouse loading and change management, with the understanding that query performance can be addressed through separate presentation layers or information marts.
Business Keys
Relationships
Descriptive Attributes
Data Vault follows an append-only pattern where records are never deleted or updated. All changes are tracked as new records with timestamps, providing complete audit trails and historical context.
Benefits:
Data Vault 2.0 uses hash keys (MD5, SHA-1) generated from business keys for primary keys. This provides consistent key sizes, eliminates key collisions, and enables parallel processing.
Every record includes metadata about its source system. Multiple satellites can exist for the same hub, each containing data from different source systems without integration conflicts.
Data Vault supports agile development by allowing teams to add new data sources, attributes, or relationships without impacting existing structures. New hubs, links, or satellites can be added independently.
Parallel Development:
Multiple teams can work on different parts of the model simultaneously without conflicts.
Initial landing zone for source data. Data loaded as-is without transformation, maintaining complete historical records.
Adds calculated fields, reference data, and business rules. Bridges the gap between raw data and business needs.
Dimensional models or aggregated views for end-user consumption. Optimized for query performance and business user understanding.