Data mart

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

A data mart is a smaller, focused piece of a data warehouse. It serves a specific department, business line, or region and contains the data that group needs to do its work. Unlike a full data warehouse, a data mart is aimed at a particular area, which can make it faster and easier for users to get the data they want. Sometimes a department owns its data mart and the hardware, software, and data, so they can control how it’s used. In other setups, shared data like customers or products means ownership isn’t strictly per department.

Warehouses and data marts exist because raw data isn’t easy to access or analyze. They are largely read-only and are designed to answer questions more quickly by providing the data in a usable form. A data mart is basically a condensed, job-focused version of a data warehouse for a specific function or region, and many departments may have their own marts to meet their needs.

A related idea is spreadmart, a pejorative term for when analysts build linked spreadsheets that become hard to maintain (often called “Excel hell”).

There are two main schools of thought about how data marts fit into the bigger picture. Inmon’s view: a dependent data mart is a subset of a central data warehouse, created for specific reasons. Tradeoffs include limited scalability, data duplication, and possible inconsistencies with other data. Kimball’s view: a data warehouse is the union of all data marts. This can lower costs and speed up development but may lead to inconsistencies in the overall warehouse, especially in large organizations. Kimball’s approach is often better for small to mid-sized companies.