What problem does data vault modelling solve, and when is its complexity justified?

The problem it solves

Integrating many source systems whose schemas change independently, while retaining a complete auditable history.

It separates business keys (hubs), relationships (links) and descriptive attributes (satellites). Because attributes are isolated in satellites with load timestamps, a source schema change adds a satellite rather than altering existing structures — so the model is additive under change, which is its central property.

And because everything is retained with load metadata, the history is complete and auditable: what was known, from which source, at what time.

When the complexity is justified

• Many source systems — tens rather than a few — with independent change cadences.
• A regulatory or audit requirement for complete lineage and history: what did we know, from where, when.
• A long-lived integration layer expected to outlast its sources.
• Frequent source changes that would otherwise force repeated remodelling.

When it is not

For a small number of stable sources feeding business intelligence, it is substantial complexity for little benefit. The join count to answer a simple question is high, the model is not understandable by analysts, and a dimensional model would serve better.

It is also not a consumption model. Data vault is an integration and history layer; business-facing marts are built on top of it, so adopting it means maintaining two models rather than one.

The costs to state plainly

• Many more tables than an equivalent dimensional model.
• Query complexity, with many joins for straightforward questions.
• Requires disciplined implementation. Partially-applied data vault has the complexity without the properties.
• Specialist knowledge, which is scarcer than dimensional modelling expertise.

The judgement

Data vault buys resilience to source change and auditable history, and pays in complexity and joins. For a large enterprise integrating many systems under audit obligation, that is a good trade. For a product company with a handful of sources, it is over-engineering that will be resented — and the complexity will be worked around rather than maintained.