About This Talk
At SmartData 2020, I presented a side-by-side comparison of Data Vault and Anchor Modeling — two highly normalized data warehousing methodologies — and proposed a hybrid model that combines the strengths of both approaches while mitigating their individual weaknesses.
Key Ideas
Data Vault’s Strengths — Hub-Link-Satellite pattern provides enterprise-wide consistency, parallel loading capability, full auditability, and strong tooling support. Its hub-centric design makes it natural for organizations that think in terms of business keys and their relationships.
Anchor Modeling’s Strengths — 6th normal form decomposition enables zero-impact schema evolution, full historization by default, and extreme flexibility. Adding new attributes or relationships never requires modifying existing structures — a powerful property for rapidly evolving data landscapes.
The Hybrid Approach — The Highly Normalized Hybrid Model (HNHM) applies Anchor Modeling’s attribute-level decomposition selectively within Data Vault’s organizational structure. Core entities use Hub-Link patterns for clarity and performance, while volatile or frequently-changing attributes use Anchor-style decomposition for flexibility.
Practical Trade-offs — The talk covered when to apply each technique: stable, well-understood domains benefit from Data Vault’s structured approach; rapidly changing or exploratory domains benefit from Anchor Modeling’s flexibility; and the hybrid model lets you optimize the trade-off at the attribute level rather than committing to one methodology for the entire warehouse.
Why It Matters
Most DWH methodology discussions present Data Vault and Anchor Modeling as competing approaches — you must choose one. The HNHM framework shows they’re complementary techniques that can coexist, giving architects more tools to address the specific characteristics of each data domain.
