Data vault
A data vault is a data warehouse modelling method that preserves source history and connects records from different sources over time. It separates stored source data from business rules and structures used to prepare data for analysis.
Example
For illustration, customer C104 has loan L208 with a reported balance of $500 on 1 March. If the source later reports $420 on 1 April, the satellite adds a dated row for the new value and keeps the earlier row rather than replacing it.
How hubs, links and satellites fit together
A hub holds one row for a business key, such as a customer number. A link records a relationship between hubs, such as a loan belonging to a customer. A satellite holds descriptive attributes and the date each version was loaded.
The raw vault stores source data with minimal transformation. The business vault adds business rules and helper structures, including point-in-time and bridge tables.
Data vaults and Kimball star schemas
A data vault and a Kimball star schema play different roles in a warehouse. The vault holds source history, and data marts or star schemas built from it are usually the structures readers query.
This pattern keeps the source history in the vault and presents data in a form that can be queried through a mart or star schema. Readers usually do not query the vault tables directly.
Questions we get asked
Who developed the Data Vault method?
Dan Linstedt developed the Data Vault method. Data Vault 2.0 is its current form.
What is Data Vault 2.0?
Data Vault 2.0 is the current form of the Data Vault method for modelling a data warehouse. Hubs hold business keys, links record relationships between them, and satellites keep dated descriptive attributes; the raw vault stores source data, while the business vault adds business rules and helper structures.
When might a lender use a data vault?
A lender might use one when it needs to preserve changing source attributes and connect records such as customers and loans. Its insert-only structure adds a satellite row when an attribute changes, keeping the earlier version.
The article on lending warehouse layers explains how raw and business vaults preserve source history, helping teams answer what a past credit decision used.