A data mart is typically defined as a subset of the contents of a data warehouse, stored within its own database. A data mart tends to contain data focused at the department level, or on a specific business area. The data can exist at both the detail and summary levels. The data mart can be populated with data taken directly from operational sources, similar to a data warehouse, or data taken from the data warehouse itself. Because the volume of data in a data mart is less than that in a data warehouse, query processing is often faster.
Characteristics of a data mart include:
Departmental or regional divisions often determine whether data marts or data warehouses are used. For example, if managers in different sales regions require data from only their region, then it can be beneficial to build data marts containing specific regional data. If regional managers require access to all the organization’s data, then a larger data warehouse is usually necessary.
Although data marts are often designed to contain data relating to a specific business function, there can be times when users need a broader level of business data. However, because this broader-level data is often only needed in summarized form, it is acceptable to store it within each data mart rather than implementing a full data warehouse.
Data warehouses can be built using a top-down or bottom-up approach. Top-down describes the process of building a data warehouse for the entire organization, containing data from multiple, heterogeneous, operational sources. The bottom-up approach describes the process of building data marts for departments, or specific business areas, and then joining them to provide the data for the entire organization. Building a data warehouse from the bottom-up, by implementing data marts, is often simpler because it is less ambitious.
A common approach to using data marts and data warehouses involves storing all detail data within the data warehouse, and summarized versions within data marts. Each data mart contains summarized data per functional split within the business, such as sales region or product group, further reducing the data volume per data mart.
Data marts can be useful additions or alternatives to the data warehouse, but issues to consider before implementation include: