Model Support for the BI Lifecycle
The business intelligence (BI) environment is built incrementally and iteratively under the auspices of an overall program. The incremental approach adds new BI capabilities, with each project being completed in a few weeks or months. The program orientation ensures that, even though the environment is constructed through a series of projects, a governance structure including standards, guidelines, and enforcement mechanisms is in place to ensure that the pieces all fit together. While each project entails understanding the needs, designing a solution, and then implementing it, the way these are done in BI projects differs from the traditional waterfall approach. BI projects are iterative in nature—the requirements often evolve as the project progresses, and managing these changes is a critical aspect of BI projects.
In addition to the unique development approach, BI environments provide business users with great flexibility in the data they access and the way they access it. From the inception of data warehousing, we’ve recognized the value of metadata both to support these business users and to capture information about the environment during its development and operation. The problem many companies have is that the metadata they collect is often fragmented and not easily accessible. The transformation maps are in spreadsheets, the definitions are in a data modeling tool, the actual transformations in the data acquisition tool, and the usage information in the end-user tool. In addition, business requirements and service level objectives, which also need to be captured and tracked, are typically in text documents. With this approach, the requirements and metadata are often not updated to reflect changes to the environment, and since some of the information is retained in multiple places, it is often inconsistent.
Most BI teams use data modeling tools to support the design of the initial data structures. To the extent practical, these modeling tools should be leveraged further to support the full BI lifecycle. This paper describes how powerful modeling tools can be invaluable in both developing and maintaining the BI environment. The approach applies equally well, regardless of the BI architecture (e.g., bus or hub-and-spoke) or methodology (e.g., iterative or agile) that is used. Regardless of the architecture and methodology,