Posted on April 4, 2013 by Sean Wang
Tags: Mythics Consulting, Business Intelligence
Data modeling is the key to success in Business Intelligence (BI). It is paramount that the process is business-centered. It starts with the clear understanding of the business, its purposes, and how the data will be used to support the business.
A data model for one line of business is hardly appropriate for another line of business. Only after a thorough analysis of an organization can a data model for the business lines be established to support the Business Intelligence process for the organization.
So, to ensure a successful Business Intelligence process, a data model should exist as follows:
- Business-Specific: As mentioned above, the data model must be developed based on the unique nature of the business, the data types and their relationships. It should represent and organize the “big data” in the particular line of the business. It should be based on how data will be searched and used in the organization.
- With Built-in “Intelligence”: The data model should include built-in “intelligence” through metadata, data dictionaries, hierarchies and inheritances. With such information, much of the “manual” inference work can be taken out of the decision process, thus, making the Business Intelligence process more efficient and effective.
- Robust: The data model should be able accurately and thoroughly represent the business, the data, and the decision process. It should reach a certain level of exhaustivity and specificity. Anything short of that will not be able to effectively support the Business Intelligence process.
- Scalable: The data model should be scalable and modular to support the ever-changing business needs of the organization. For instance, a subset of the data model could be adapted to more efficiently support a small business unit within the organization. Or, the data model could be easily expanded to support future business expansion, or as new data becomes available.
- Implementable: The data model should be easily implemented, either through in-house development, or through commercial tools, such as Oracle’s BI Publisher.
Bottom-line: Understand the business and data before modeling the data for Business Intelligence.