Fundamentally, a dimensional model deals with the measurement of business processes. It describes how a business process is evaluated, and can be used to frame questions about a process. In this respect, it speaks clearly to the business users of the data warehouse.
A dimensional model also has technical implications. Its definition determines the data sources that must be integrated, how information must be cleansed or standardized, and what queries or reports can be built. In this respect, it speaks clearly to the developers of the data warehouse.
These business and technical characteristics of the dimensional model make it an ideal focal point for managing the entire data warehouse life cycle. A dimensional model can serve as the basis for a shared understanding of warehouse strategy. From a business perspective, it imparts a clear understanding of functional capability; from an I.T. perspective, it supports a clear understanding of technical activity.
- Warehouse Strategy It is well understood that planning a dimensional model at an enterprise level can enable the incremental implementation of subject area applications. For a Kimball-style dimensional data warehouse, this process is critical to ensuring that stove-pipes are not developed.
But a high-level dimensional model has additional value, even if you are not building a dimensional data warehouse. Because it provides a clear framework to describe functional capability and technical work-units, a dimensional model is an effective way to plan and document an enterprise data warehouse architecture.
I use the dimensional model as the central focus of data warehouse strategic plans. It is understood by business and technical constituents, bringing them together with a shared understanding of scope, functionality and technical effort for each subject area. It clearly conveys functionality, while at the same time allowing development activity to be quantified.
- Requirements Definition In the same way, a dimensional model is an ideal way to capture requirements—whether at a strategic level, or for a subject-area implementation (or data mart.) Report requirements, data requirements, and loading requirements can all be expressed in dimensional terms. This makes each requirement understandable to a variety of audiences, and allows their dependencies to be easily cross-referenced.
I use a dimensional format to capture requirements and develop specifications for deliverables such as reports, applications, ETL routines, and, of course, schema design.
- Project Prioritization A dimensional model can be cross referenced with business priorities, report functionality, data availability, load requirements, and several other factors that drive a development roadmap. As a common ground for business and technical interaction, it is an invaluable tool.
Once priorities are set, the dimensional framework can be used to describe them. By segmenting a dimensional model into a set of sequenced projects, it clearly links both functionality and technical effort to the calendar. At the same time, it enables analysis of resource requirements that will be required.
- Manging Scope The dimensional model is an ideal way to define and manage the scope of projects. I use dimensional terms to describe project objectives, what is considered in and out of project scope, and how change requests will be evaluated by project leadership.
This is particularly useful for implementation projects that employ an "iterative" build process, where scope-creep is an ever-present possibility. By linking project scope to dimensional characteristics (such as grain, sources, or attribution), physical design can be enhanced through iterations without allowing the project to spiral out of control.
© 2007 Chris Adamson