Databases Reference
In-Depth Information
It is common to continue to find attributes to add to dimensions that have
already been drafted. Simply add the new attribute to the diagram and go
back to what you were working on.
The objective is to get a first draft of all of the dimensions, not to get a perfect
version before moving on to the next dimension. If the group has spent more
than ten or fifteen minutes discussing or debating a set of data elements, then
log the questions and move on. These issues will be reviewed and resolved
later. If the team can't figure out what to do fairly quickly, then other people
may be needed to provide more information or it may be too early in the
modeling process. Later, it may seem immediately obvious how to handle a
data element when the rest of the model is fleshed out.
Business dimensions should include attributes that contain complete values
or descriptions. It is not acceptable to include only codes in the dimension.
These codes are often well understood by people who currently access this
data because more complete descriptions are not readily available. Over time,
people have gotten used to these codes and know their meaning. However, it
is difficult for new people to learn them. Many of the current users of the data
have forgotten what it was like when they were trying to figure the data out.
The objective is to empower people who have not had access to the data in the
past. These people are not likely to know what these codes stand for. Useful
business names and descriptions should be included in the dimensions. Codes
can also be included in the dimensions as needed.
To help create a meaningful visual representation of the dimension, place
the lowest-level attribute at the bottom center of the diagram. No other ele-
ments should appear lower than this element. The main or primary hierarchies
should be placed up the center of the diagram. This shows their importance
and the eye is drawn quickly to them. If other attributes describe similar or
related things, then these should be diagrammed together. For example, if
the Education Level Completion Date is needed, it should be placed near the
Education Level attribute. Keep the attributes in the same general location
with each revision of the model. This helps reduce confusion and keeps people
from thinking that attributes have been removed. If there is a need to clean up
a dimension, do this all at once, and if possible prior to presenting the model
to a wide group of people.
Drafting the Initial Fact Groups
Once the first cut of all the dimensions is completed, it is time to take a look
at the facts. Start with facts that are straightforward and well understood.
Locate the transaction tables in the source systems where these facts are
captured. By studying the context of these facts in the source system, you can
see which dimensions will apply. The source system dependencies and keys
are helpful to understand the data itself. After you understand the facts, you
