A data warehouse is a subject-oriented application. Subjects are the abstraction of data that is aggregated, sorted, and analyzed for future use. In addition to horizontal layering, the model design of the data warehouse requires you to divide data domains in the vertical direction based on business conditions. A data domain is a collection of closely related data subjects and a conceptual classification of business objects. It aims to provide convenience for you to manage and apply data.
Divide data domains
Before dividing data domains, you need to read the design documents, data dictionaries, and data model design documents of different source systems and study the physical data models generated through reverse engineering. Then, you can merge subject domains that come from multiple source systems to build and sort out the data domains of the entire enterprise.
A data domain is an abstract collection of business processes or dimensions based on business analysis. To ensure the vitality of the data warehouse, data domains need to be extracted and maintained at regular intervals. When dividing data domains, you need to ensure that they meet all the current business requirements. If new business requirements are generated, the existing data domains are still applicable or can be extended to meet new requirements. We recommend that you divide data domains after you analyze the business activities of each business unit in a business survey.
|Data domain||Business process|
|Member shop domain||Registration, logon, decoration, opening, and closing|
|Product domain||Posting, putting on the shelf, taking off the shelf, and re-posting|
|Log domain||Exposure, browsing, and click|
|Transaction domain||Order placement, payment, shipping, and receipt confirmation|
|Service domain||Adding to favorites, visit, training, and coupon collection|
|Procurement domain||Product procurement and supply chain management|