From Data Vault to Model-Driven Engineering — How the foundations laid by Hans Hultgren’s Ensemble Modeling still guide my approach today.
Summary
Hans Hultgren’s work on Ensemble Modeling and Data Vault has deeply shaped my thinking around agile, scalable, business-oriented data architectures.
My direct collaboration with Hans during a Data Vault implementation at Rabobank International, as well as my certification and participation in DVEE Consortium meetings, solidified the foundational principles I use today: separation of concerns, metadata-driven design, and agile evolution of data models.
Meeting Hans Hultgren and Discovering Ensemble Modeling
In 2015, I had the privilege of earning my Data Vault 2.0 certification directly from Hans Hultgren in Stockholm. At the time, I was working on a challenging project at Rabobank International, tasked with improving the design and performance of an existing data warehouse.
Collaborating with Hans went far beyond theory: we applied his Ensemble Modeling approach hands-on to refactor and optimize the warehouse. This real-world application — focused on separating business keys, context, and relationships — allowed me to deeply internalize the principles behind Data Vault and business-oriented modeling.
Key Principles I Learned from Hans Hultgren
Hans’ book Modeling the Agile Data Warehouse with Data Vault and his workshops emphasized several transformative ideas:
1. Business Keys as the Anchor of Integration
- Business entities should be modeled around real-world, stable identifiers
- Avoid technical keys that break cross-system consistency
- Treat business keys as metadata — discoverable, explainable, documented
2. Separation of Concerns
- Separate the what (Hubs), the descriptive context (Satellites), and the relationships (Links)
- This allows agility, auditability, and incremental delivery
- Great modeling is about organizing meaning, not just organizing data
3. Model Agility and Evolution
- Design with change in mind: new attributes go in new Satellites
- New relationships can be added without breaking the model
- This is crucial for long-lived architectures that evolve with the business
4. Ensemble Modeling = Data Vault for the Broader Modeling Community
- Ensemble Modeling generalizes Data Vault concepts: it’s about structuring business data as ensembles (Hubs + Satellites + Links + sometimes Decouplers or Ties)
- It brings together Data Vault, Anchor Modeling, Focal Modeling, and other variants under shared principles
How It Influences My Current Work
Years later, these principles still guide how I approach model-driven data engineering, business-friendly mapping, and metadata-first data architecture.
🧱 Model-Driven Engineering
While my thinking was initially influenced by the classic separation of Hubs, Satellites, and Links, my recent work has evolved beyond strict Data Vault implementations. In my current approach, I no longer physically implement Hubs, Links, and Satellites as rigid structures.
My focus is on building business-aligned integration models that retain the clarity and agility of Ensemble Modeling but avoid over-engineering. This shift has allowed me to design platforms that are easier to explain, maintain, and scale across teams — without depending on a handful of Data Vault experts.
I shared this evolution in my thinking in the article: “Lessons from Data Vault: Principles Without the Dogma”. There, I argue for applying the valuable principles of Data Vault — like separation of business keys, context, and relationships — without being constrained by its technical dogma. This helps avoid complexity, reduces the risk of single-person dependencies, and improves team-wide understanding and ownership.
🗂️ Business-Oriented Modeling
When designing integration layers or conceptual models, I think in ensembles. Even outside strict Data Vault patterns, I look for business keys, temporal context, and clearly modeled relationships.
📐 Metadata-Driven Mapping
Hans taught me to see the model as a knowledge artifact, not just a diagram. I use metadata schemas to define mappings, transformations, historization, and lineage — everything that reflects the business intent of data.
The Impact of the DVEE Consortium
Beyond my work with Hans, I participated in several DVEE Consortium meetings — multi-day events where modelers from different organizations explored ensemble modeling patterns, governance, automation, and scaling models in complex organizations.
Those sessions reinforced that we are not just building data warehouses — we’re building shared language systems. Ensemble Modeling gave us the tools to do that cleanly, iteratively, and transparently.
Where Hans Hultgren’s Thinking Still Leads the Way
Whether you’re modeling a small domain or a full enterprise data platform, Hans’ principles remain:
- Model the business, not the system
- Separate logic cleanly to support change
- Think in ensembles, not monoliths
- Use metadata to automate and govern
And above all:
“Think of the model as a map of meaning, not a mirror of a system.” — Hans Hultgren
Learn More
- 🎓 Hans Hultgren’s Ensemble Modeling Course (PDF)
- 📘 Modeling the Agile Data Warehouse with Data Vault
- 🎧 AgileData Podcast — The Patterns of Data Vault with Hans Hultgren
- 💡 Data Vault Certification Info