Jaco van der Laan

The Historical Staging Area: How Roelant Vos Paved the Way Beyond Classical Data Vault

21/09/2025  ·  Data Architecture

article base/historisation-temporal
‹ ›

The Historical Staging Area- How Roelant Vos Paved the Way Beyond Classical Data Vault

Revisiting Roelant Vos’s Virtual Data Warehousing training and HSA concept, and how it anticipates bi-temporal architectures beyond classical Data Vault.

Summary

In 2019, during Roelant Vos’s Virtual Data Warehousing training, I first encountered the concept of the Historical Staging Area (HSA). Unlike the traditional Persistent Staging Area (PSA) proposed by Dan Linstedt in Data Vault methodology, Vos’s HSA was more than just a landing zone for raw history. It combined historization, queryable storage, and views that layered business meaning on top.

By questioning the literal application of Data Vault and introducing HSA as a more flexible construct, Vos effectively anticipated some of the directions I later explored with bi-temporal SCD2 architectures (Dual SCD2: The Foundation for True History in Data Warehousing, From Historization to Exposure). Both approaches aim at the same outcome: making historized data directly usable and transparent without endless remodeling.

The Historical Staging Area (HSA)

Roelant Vos introduced the HSA as a place where all changes from source systems are preserved historically — but with a difference:

  • Instead of simply storing unchanged copies (like PSA), the HSA keeps historical versions aligned to business keys.
  • On top of this raw historization, views are defined to make the data consumable, effectively turning staging into a queryable, governed layer.

This was a subtle but profound shift. Rather than forcing every query to join through hubs, links, and satellites, Vos demonstrated how the HSA could serve as a virtual integration layer.

Questioning Classical Data Vault

In his blog (roelantvos.com/blog) and the Virtual Data Warehousing course, Vos openly questioned whether implementing Data Vault literally as prescribed by Dan Linstedt was always the best option.

Classical Data Vault:

  • Hubs, Links, Satellites persist all raw data.
  • Requires heavy modeling effort upfront.
  • Usability often requires another downstream model (star schemas, marts).

Vos’s HSA critique:

  • Why materialize everything if a view-based approach could expose historized data directly?
  • Why chain multiple layers if you can query history with views on the staging layer?
  • Why bind yourself strictly to patterns-as-physical-tables if virtualization can deliver the same semantics?

In this sense, the HSA was a step away from a rigid “by the book” Data Vault and toward a more agile, model-driven mindset.

Alignment with Bi-Temporal SCD2

My own work on bi-temporal SCD2 (Dual SCD2, Snapshots, and PIT Tables: A Technical Deep Dive, Why Surrogate Keys Are Essential for Dual SCD2 + PIT Tables) takes this thinking further:

  • Instead of persisting raw change data in satellites, we structure it as bi-temporal SCD2 tables (system time + business time).
  • On top of this, we define exposure views (Designing Exposure Views) that give consumers a consistent view of entities as-of any point in time.
  • The goal is the same as Vos’s HSA: usable history without unnecessary complexity.

Both approaches emphasize:

  • Historization as a foundation, not an afterthought.
  • Views as the main abstraction layer, not just physical models.
  • Business-shaped usability from the start.

Where Vos’s HSA blurred the line between staging and integration, my bi-temporal SCD2 approach formalizes that idea into a reproducible pattern.

A Personal Reflection

I had the privilege of attending Roelant Vos’s training in 2019 and meeting him in person. His way of thinking — always challenging orthodoxy, always questioning “why not simpler?” — was inspiring.

For me, the HSA was an early signpost that Data Vault would evolve. Today, as more practitioners embrace metadata-driven, bi-temporal, and virtualization-based architectures, Vos’s thinking feels more relevant than ever.

The HSA Manifesto

  1. History First — every change preserved, nothing lost.
  2. Queryable by Design — staging is directly usable.
  3. Views, Not Copies — historization exposed via virtualization.
  4. Business Keys Anchor History — historization aligned with stable identifiers.
  5. Integration at Staging — blur the line between staging and integration.
  6. Agility over Orthodoxy — don’t follow Data Vault “by the book.”
  7. Usability is the Goal — history must be business-usable, not just stored.

Conclusion

The Historical Staging Area was never just about staging — it was about rethinking how we preserve and use history. Roelant Vos showed us that literal Data Vault implementation is not sacred, and that sometimes views + historization can deliver more value than strict table patterns.

In many ways, his HSA anticipated the shift toward bi-temporal, model-driven data engineering — the same direction that now defines my own work on dual SCD2, PIT tables, and exposure views.

📝 Summary

Roelant Vos’s Historical Staging Area (HSA) questioned classical Data Vault implementation and introduced a view-based, historized staging approach that blurred the line between staging and integration.

His ideas resonate strongly with modern bi-temporal SCD2 architectures, where historization and usability are combined into a single layer.