A data mart by definition is a repository of integrated data from different data sources (internal operational data or external data) designed to serve a specific business group. “A data mart usually is organized as one-dimensional model as a star-schema (OLAP cube) made of a fact table and multiple dimension tables.”

Data marts are per se small warehouses designed for a strategic business unit, typically consisting of a single subject area, such as finance, marketing, operations, and so forth. Data marts have the data warehouse concept of online sales, inventory, and other crucial business data that has been gathered from transaction processing systems of different businesses to departments within an organization (Westerman, 2001; Stair & Reynolds, 2006).

Historically, data marts were developed as ‘prototypes’ often to serve as a proof of concept prior to investing the resources needed to develop a comprehensive enterprise data warehouse. Today, a data mart may or may not obtain its data from data warehouse since emphasis is put on easy access to information to meet specific business needs.

Multiple data marts can exit in a single business organization, each assigned to a business unit (group or department) that requires access to detailed data. These small business units can build their own decision support systems without relying on a centralized information technology department. As a result, data marts are advantageous in terms of high availability, scalability, performance, and reliability.


References:

Stair, R. & Reynolds, G. (2006). Principles of Information Systems.
 7th Edition. Thompson Course technology


Rob, P. & Coronel, C. (2004). Database Systems: Design,
Implementation, & Management. 6th Edition. Thompson Course
technology

Kimball, R. & Ross, M. (2002). The data warehouse toolkit: The
complete guide to Dimensional Modeling. 2nd Edition, illustrated.
Pub by John Wiley and Sons