Data Vault is an enterprise modeling approach for integrating many sources with auditability and agility. Core building blocks:
Hub
- Unique business keys for a core entity (customer, product, order)
- Thin: hash key, business key, load timestamp, source
Link
- Relationships / transactions between hubs (customer placed order, order contains product)
- Thin: hash of related hub keys, load metadata
Satellite
- Descriptive attributes and history hanging off a hub or link
- Where change tracking lives (often insert-only history)
hub_customer hub_order
\ /
\ /
link_customer_order
|
sat_order_details (status, amounts, effective timing)Mental model
Hubs = nouns, links = verbs/relationships, satellites = adjectives and history.
Why teams use it
- Ingest raw source truth quickly without forcing a star on day one
- Preserve lineage and multi-source conflicts
- Build star marts later on top of the vault
Interview tip
> "Data Vault separates business keys (hubs), relationships (links), and attributes over time (satellites). It is an integration layer, often under BI-facing stars."