Abstract

In this article, I share my definition of an enterprise data warehouse (which I believe is commonly agreed upon by industry practitioners).

Definition

An enterprise data warehouse (EDW) is a collection of data assets that support a business's enterprise-level decision-making processes.

Traditionally, an EDW is:

· Subject-oriented: Organized by subjects relevant to the business.

· Integrated: Combines data from conflicting and disparate sources into useful data assets.

· Non-volatile: Provides consistent results across system time.

· Time-variant: Represents reality across business time.

I remember this with the unoriginal and admittedly unintuitive acronym of SINT.

The industry-accepted goals of a data warehouse include:

· Data from across the enterprise (and beyond) is integrated into a single version of the truth.

· The data is organized in a subject-oriented manner that is meaningful to the org.

· Data values are captured as they change over time to support time-based analysis.

· The history of the data changes is non-volatile to maintain consistency in reporting.

· Data is maintained at a sufficiently granular form to enable any future reporting needs.

· Data lineage is clear and auditable.

· The architecture is scalable.

· Changes to the data warehouse can be made in an agile manner.

· Metadata is available to aid consumers of the data.