Data Mart

Data Mart is a subset of a data warehouse focused on a single business process, a functional area (sales, marketing, finance) or a department. It typically holds one or a few fact tables and the dimensions related to them.

Why data marts exist #

In real organizations, a complete enterprise DWH takes years to build. Data marts are a pragmatic compromise: build first the piece a department needs now (e.g. a sales data mart for marketing) and integrate it with the others later. This is Kimball’s bottom-up approach.

The divergence risk #

When multiple data marts are built independently by individual departments — often with different BI tools, on different source systems, on different timelines — the risk is that “customer” ends up meaning three different things across three data marts. Totals don’t match, cross-department analysis becomes impossible or slow, and the CFO ends up with three versions of the truth.

Conformed vs independent data marts #

The critical difference is whether the data mart shares conformed dimensions or not:

• Conformed data marts (Kimball): share conformed dimensions (customer, product, time, geography) and can therefore be queried together consistently
• Independent data marts: built without shared governance, drift over time and generate the classic “three versions of the truth” problems

The bus matrix is the design tool that prevents the second scenario.

When it makes sense #

A data mart makes sense when:

• The functional scope is well-defined (one process, one department)
• Conformed dimensions are already available or will be built alongside
• The cost of a full enterprise DWH isn’t justified
• You need fast time-to-value for a specific use case

It doesn’t make sense as a “permanent isolated solution”: either it’s the first piece of an integrated strategy, or it turns into technical debt within a few years.