pattern

Hub and Satellite

Separating stable business keys from their changing attributes and from their relationships, so each can be loaded independently and kept forever.

Data Vault splits what dimensional modelling combines. A hub holds only a business key and its metadata — the list of order numbers that have ever existed. A link holds a relationship between hubs. A satellite holds descriptive attributes for a hub or link, with a load timestamp, and is append-only.

Three properties follow. Loads are independent and therefore highly parallel, and adding a new source means adding satellites without touching what exists. Everything is auditable, because nothing is ever updated — the satellite retains every version with the time it arrived. And the model is resilient to source change, since a restructured source becomes another satellite rather than a migration.

The cost is that it is not queryable by a business user. A simple question spans many joins, so a dimensional layer is built on top for consumption, and the organisation now maintains two models.

That trade is worth it where auditability is a regulatory requirement and sources change often — which describes a lot of banking and insurance, and very little else. Adopting it for a stable three-source estate buys complexity and no benefit.