Ensemble Modeling Evolved- How Hans Hultgren and Remco Broekmans Shape the Future of Model-Driven Data Engineering
From unified decomposition to business-friendly workshops, Ensemble Modeling’s latest insights strengthen the bridge between data modeling, communication, and automation in our methodology.
Why Ensemble Modeling Still Matters
When I earned my Data Vault certification directly from Hans Hultgren back in 2015, one thing struck me: Data Vault wasn’t just another modeling technique — it was a way of thinking. By separating business keys, relationships, and context (Hubs, Links, Satellites), Hans introduced the principle of unified decomposition: break concepts into atomic, auditable parts while keeping them tied together.
That balance of agility, auditability, and integration became the foundation of Ensemble Modeling. And it’s more relevant today than ever, as we rethink how business and IT collaborate in data architecture.
Hans Hultgren’s Legacy: Agility Through Structure
Hans’s Modeling the Agile Data Warehouse with Data Vault laid out why traditional approaches (3NF or Dimensional) struggle at EDW scale:
- 3NF optimizes for operations, but not historization.
- Dimensional models optimize for reporting, but not integration.
- Data Vault strikes a middle ground — built for historization, auditability, and agility.
Key takeaways:
- All relationships are M:M: The warehouse should reflect reality, not rules.
- DW Controlled Attributes are “untouchables”: Load date, record source, sequence IDs provide trust and traceability.
- Concept Constellations show how ensembles aggregate into richer business concepts.
For our Model-Driven Data Engineering (MDDE), Hans’s work reinforces the structural backbone: historization, separation of concerns, and clear lineage.
Remco Broekmans’s ELM Approach: From Stories to Solutions
Where Hans provided the structural grammar, Remco Broekmans has focused on language and communication. His new book From Stories to Solutions introduces the Ensemble Logical Model (ELM), a guided approach to bridging business and IT.
At its core, ELM is about solving the communication gap: business users speak in stories, IT speaks in models. ELM turns business conversations into logical ensembles without distortion.
The Three Phases of ELM Workshops
- Align: Gather Core Business Concepts (CBCs) and understand business vocabulary.
- Refine: Explore relationships, directions, and grain.
- Design: Build the Ensemble Logical Model collaboratively.
Six Key Artifacts
- CBC-List: Captures business concepts in natural terms.
- CBC-Canvas: Categorizes concepts (Event, Person, Place, Thing, Other).
- CBC-Form: Defines concepts with context and identifiers.
- Event-Canvas: Maps events and their related CBCs.
- NBR-Matrix & NBR-Form: Capture and detail natural business relations.
Together, these artifacts guide the conversation — making models transparent, repeatable, and aligned with business meaning.
Lessons from ELM: Communication Before Modeling
Remco emphasizes: “This is not about data modeling. It’s about capturing stories, communication, and conversation.”
- Facilitation matters: Avoid jargon like “logical data model.” Instead, ask: “How do you talk about your work?”
- Categorization prevents confusion: Events vs. Persons vs. Things reduces synonym clashes.
- Repetition builds trust: Revisiting workshops ensures alignment even when stakeholders change.
- AI as an assistant, not a replacement: GPTs/LLMs can help propose CBCs, generate draft canvases, or flag ambiguities, but facilitation remains human-driven.
Connecting to Our Methodology: MDDE in Practice
Where does this fit into our dual / bi-temporal SCD2 and view-first architecture?
- Unified Decomposition Meets Bitemporal SCD2
We don’t implement Hubs, Links, Satellites directly. Instead, our historization is based on 3NF business models extended with bitemporal SCD2 (no updates, only new versions and logical deletes). This preserves history while keeping the model aligned with business semantics. - ELM Workshops Feed Metadata
CBC-Lists and NBR-Forms can directly populate our metadata repository. That metadata then drives draw.io or Miro diagrams, SQL generation, and exposure views. - Business-Friendly Mapping Enhanced
By grounding mapping steps in CBC categories and natural business relations, we add another layer of clarity. Business conversations → mappings → executable SQL. - Automation with Guardrails
Metadata captured via ELM artifacts can flow into scripts for automated diagram generation, schema creation, and view layers — but always anchored in business terminology.
The Synergy: Hans + Remco + MDDE
- Hans Hultgren: Structural rigor (decomposition, historization, auditability).
- Remco Broekmans: Communication frameworks (stories, workshops, CBC artifacts).
- Model-Driven Data Engineering: Automation and delivery (metadata, dual SCD2, exposure views).
Together, they create a virtuous cycle:
👉 Business stories → Ensembles → Metadata → Automated Engineering → Business exposure
This cycle is exactly what modern data teams need: agile, communicative, and reliable.
Why This Matters Now
The modern data ecosystem is noisy: lakehouses, data meshes, semantic layers. Amidst the noise, Ensemble Modeling and MDDE offer stability:
- For business: clarity, communication, and trust in definitions.
- For engineers: automation, historization, and governance.
- For architects: a framework that scales without losing agility.
By combining Hans’s structural foundation and Remco’s communication practices, we can make MDDE not just an engineering framework but a business-first architecture for the data age.
References
- Hans Hultgren — Modeling the Agile Data Warehouse with Data Vault
- Remco Broekmans — From Stories to Solutions (ELM Approach)
- Genesee Academy — Training and certification in Ensemble Modeling