Jaco van der Laan

Universal Metadata Models: From a Blueprint in 2014 to Modern Metadata Architectures in 2025

21/09/2025  ·  Data Modeling

article base/metadata-models
‹ ›

Universal Metadata Models- From a Blueprint in 2014 to Modern Metadata Architectures in 2025

How I used a 2004 classic to design a metadata model for Rabobank — and how its lessons still guide metadata solutions today.

Summary

In 2014, I used Universal Meta Data Models as a blueprint to design the metadata model for Rabobank’s ETL framework in a Data Vault warehouse.

The book defines a Managed Metadata Environment (MME), with universal models for technical metadata, messaging, IT portfolio, and stewardship — unified in a Universal Meta Model (UMM).

Its test still applies: if you can generate lineage, impact, rules, and quality reports, your metadata is effective.

Since 2004, new standards (ISO 11179, DCAT v3, FAIR) and platforms (Egeria, DataHub, OpenMetadata, Atlas, OpenLineage, Unity Catalog) have operationalized these concepts.

My current goal is to blend the timeless UMM with modern tools to power governance, lineage, contracts, and quality in the modern data stack.

Introduction: My Journey with Metadata

In 2014, while working at Rabobank International on a Risk & Finance data warehouse program, we were building a Data Vault–based warehouse with a custom ETL framework.

One of my responsibilities was to design a metadata model to support ETL code generation and orchestration. It was a big challenge — we needed something that wasn’t just technical plumbing, but an approach that supported governance, lineage, and business rules.

That’s when I bought Universal Meta Data Models (2004) by David Marco and Michael Jennings.

The book quickly became my blueprint. Its universal models, example reports, and structured methodology gave me a practical way to think about metadata not as an afterthought, but as the backbone of our ETL framework. Looking back, this decision shaped much of how I still think about metadata today.

What the Book Taught Me

Marco & Jennings make a crucial distinction: a Managed Metadata Environment (MME) is not just a repository — it’s an enterprise-wide capability.

Each chapter follows a clear pattern: purpose → assumptions → subject areas → sample reports → summary. That format was a game-changer for me when documenting and communicating with teams at Rabobank.

Here’s how the book breaks down metadata:

1. Enterprise Systems (Technical Metadata)

  • Subject areas: Data packaging, data movement, transformations, profiling, lineage.
  • Reports: Lineage diagrams, profiling exceptions, impact analysis, data quality reports.

2. XML, Messaging & Business Transactions

  • Subject areas: XML schemas, DTDs, transformations (XSLT), messaging flows.
  • Reports: XML usage, import tracking, process impact analysis, transaction flows.

3. IT Portfolio Management

  • Subject areas: Services, software, hardware, networks, projects, data quality.
  • Reports: Hardware/software obsolescence, CRUD matrices, project success rates, exception tracking.

4. Business Rules, Business Metadata & Stewardship

  • Subject areas: Business terms, rule catalogs, stewardship roles, issues/decisions.
  • Reports: Business rules by category, data groups by subject area, open issues.

5. The Complete Universal Meta Model (UMM)

  • A consolidated enterprise model with entities for systems, projects, people, data structures, rules, and processes.
  • Includes a Data Quality Metadata Mart and OLAP implementation for reporting.

💡 Lesson learned at Rabobank: the real proof of a metadata model is whether you can generate those reports. That became my guiding test: if lineage, impact analysis, and quality dashboards worked, we knew the model was good enough.

What’s Changed Since 2004

When Marco & Jennings wrote their book, XML was cutting-edge. Today, metadata has expanded into event-driven platforms, cloud lakehouses, and open-source catalogs.

Here’s what we now build on top of their ideas:

Modern Standards

  • ISO/IEC 11179: Still the gold standard for precise metadata registries.
  • W3C DCAT v3: Web standard for dataset catalogs and FAIR alignment.
  • FAIR principles: A north star for making metadata useful (Findable, Accessible, Interoperable, Reusable).

Open Source Platforms

  • Egeria: Open metadata types, APIs, and federation.
  • LinkedIn DataHub: Stream-based metadata propagation.
  • OpenMetadata: Fast-growing platform with governance and lineage built in.
  • Apache Atlas: Classic for classifications, policies, and lineage.
  • OpenLineage + Marquez: Lineage event standards for Airflow, Spark, dbt.

Enterprise Integrations

  • Databricks Unity Catalog: Unified access and lineage for the lakehouse.
  • Data Contracts: Schema + SLA agreements between producers and consumers, bringing accountability to data pipelines.

My Goal Today

Revisiting the book now, my goal is to design a modern metadata model that merges its timeless UMM with today’s standards and tools.

Core entities

  • DataAsset, Schema, Field → technical metadata.
  • BusinessTerm → linked glossary and definitions.
  • Transformation, Job, Run → operational metadata and lineage (via OpenLineage).
  • Policy, Classification, Contract → governance and accountability.
  • QualityCheck & Result → automated data quality scorecards.
  • Organization, Person, Steward → ownership and stewardship.

Relationships

  • Ownership: Steward → DataAsset / BusinessTerm / Policy.
  • Provenance: Run consumes DataAsset A → produces DataAsset B.
  • Governance: Policy/Contract → applied to DataAsset.
  • Quality: Check → DataAsset → Result (time-series).

This model doesn’t just live in documentation — it becomes operational, powering catalogs, APIs, dashboards, and governance workflows.

Implementation Blueprint (Databricks + Airflow Example)

Here’s how I’d put it into practice today:

  1. Catalog foundation: Unity Catalog as the authoritative schema and lineage layer.
  2. Enterprise catalog: Egeria, DataHub, or OpenMetadata for governance, glossary, policies.
  3. Lineage plumbing: OpenLineage events from Airflow and Spark.
  4. Glossary & rules: BusinessTerms tied to fields, plus rule catalogs.
  5. Contracts: Introduce contracts on high-value domains, gate schema changes via CI.
  6. Reporting tests: Recreate the book’s sample reports as dashboards. If they break, metadata coverage is incomplete.

Closing Thoughts

When I bought this book over a decade ago, I didn’t know I would still be applying its lessons today. But solid architectural patterns don’t age — they evolve.

By combining the Universal Meta Data Models framework with modern standards and platforms, we can finally realize what Marco & Jennings envisioned: a metadata environment that is not just a repository, but a living ecosystem of trust, automation, and governance.

And that is exactly the model I continue to pursue today.

✍️ This article is Part 1 of the series “Why Elephants Matter in Data Modeling”, where I explore how John Giles’ ideas, Len Silverston’s Universal Data Models, and my own methodology come together in a business-first, model-driven approach.

Series overview

  • Part 1: Why Elephants Matter in Data Modeling
  • Part 2: Patterns + Universals: The Foundations of Reuse
  • Part 3: From Ontology to Enterprise View
  • Part 4: Borrowing from Data Vault Without Building One
  • Part 5: Model-Driven Engineering in Practice
  • Part 6: Business-Friendly Mapping as a Collaboration Tool
  • Part 7: The Four Pillars of Business-Centered Data Architecture