Databases Reference
In-Depth Information
Characteristics of the Model across the Enterprise
As a data warehouse grows over time or a broad model is developed up
front, there can be many dimensions and fact groups. It is not unheard of to
have thirty or forty different dimensions as you look across an enterprise. You
are likely to find is that a core set of dimensions is used repeatedly. These
must be conformed! The other dimensions are typically used for fact groups
that pertain to only one part of the business. Using an education example,
the school district, student, and teacher dimensions will be used for many
fact groups. Dimensions about student disciplinary actions, extracurricular
activities, or teacher licensure pertain only to some of the fact groups.
TIP
While enterprise-wide models may have over thirty dimensions and fact
groups, keep in mind that any single fact group will have between five and fifteen
dimensions.
Many organizations are already involved in data warehousing, perhaps
even having multiple data warehouses. These have often been built without
using conformed dimensions. By bringing the dimensions and facts together
into a single model, this can be a good start toward understanding what will
be needed to integrate these stovepipe databases.
Many enterprise-level dimensional models can be complex, because of the
breadth that is included. Most users have an interest only in a subset of this
model. The scope of what is presented to each group of users can be tailored
to meet its own specific needs. Very few people in the organization may see or
have a need to use the entire model.
ENTERPRISE DESIGN GUIDELINES
■
Design broadly.
■
Don't try to tackle all data for the enterprise at once.
■
Design at the lowest level of detail that is captured.
■
Work with multiple user groups who use the same data.
■
Always consider an enterprise viewpoint—who else could use this data?
■
Always design and plan for conformed dimensions.
■
The dimensional model can be implemented in several increments.
■
Each increment should deliver business value.
■
Review and update the model with each increment.
