|
Return to Newsletter Contents...
Don't Care Yet - Now You Should
by: Troy Gottfried, Sr. Consultant ,
CRCP, BECP
In the first installment of
this series, we discussed the advantages of utilizing the Unified Dimensional
Model (UDM) in Microsoft SQL Server 2005 Analysis Services.
We saw how hierarchies, categories, and built-in time dimensions help to
save development time and capitalize on the time invested in implementing
Microsoft’s new semantic layer. This
issue will continue the discussion, focusing on Key Performance Indicators
(KPIs), Perspectives, Attribute Semantics, and cube write-back capabilities.
Key Performance Indicators, or
KPIs, offer decision-makers a tool to ‘manage by exception.’
It is worth the effort to step back and think about exactly what that
means for a moment before continuing with the discussion.
Managing by exception is a concept built around the premise that we can
capture values that are integral to the business and display them in a format
easy to comprehend. Furthermore, it
suggests that by viewing the visualization of those metrics, the manager has
at their disposal the information to decide whether an action needs to be taken.
The UDM covers all of the
above assumptions, and more, with regards to KPIs.
First, KPIs can consist of up to four expressions:
(1) Actual Value, (2) Goal Value, (3) Status, and (4) Trend.
Of these four expressions, the client tools we use to display the
information will show values and goals as numbers, with status and trend as
visualizations. Additionally, the
UDM allows these values to be on-demand, or even ‘cached’ on a periodic basis
(scheduled to refresh on a time interval).
Because these are all stored in the UDM, they can be reused in a variety
of applications, always ultimately sourcing the same data.
This provides continuity and ‘one version of the truth’ regardless of who
is viewing the KPI.
Perspectives also simplify the
process of analyzing information with the UDM.
This is one of my personal favorites.
For many years, we have used views built with SQL to secure and display
information to users to help simplify the information they see.
With OLAP (On-Line Analytical Processing) tools, it is easy to get lost
quickly when analyzing the information contained in the cube.
The number of measures and the number of dimensions used within any
multidimensional data structure can quickly become overwhelming.
Perspectives are essentially like views for multidimensional data
structures. Subsets of the model can
be presented to individuals so that they are only seeing the information that is
relevant to their decision-making process.
Perhaps a more nebulously
named, yet equally important feature in the UDM is the use of Attribute
Semantics. Attribute Semantics allow
each attribute to be assigned both a key and a name.
Where is this useful? We can
look to the Simpsons for help. We
know that the Simpson family lives in a town called Springfield.
Now, which Springfield it is, we don’t know.
However, each city in our model will have a unique ID associated with it.
We can apply both the unique ID as the key, and Springfield as the name
to each of these cities, ensuring that all Springfields in our analysis are
accounted for, while representing them with the appropriate name.
We can also decide on how to
order our attributes, without being confined to the common A to Z dogma we have
gotten frustrated with so often. For
example, we can choose to always order our months of the year as January,
February, March, etc… without having to display the 1, 2, and 3.
Along these lines, we also have at our disposal another useful tool (and
nebulous name) called "discretization".
If we think of the root of that made-up word, we can pick out the word
discrete. Sometimes, it is not all
that useful to see all of the discrete values.
Rather, it might be more useful to arrange those values into ranges.
In this instance, we could take the following values:
1034.23, 343.26, 928.38, 563.01 and move them into ranges such as:
<500, 500 – 800, 800.01 – 1000, >1000.
Such is Discretization.
Finally, we come to cube
write-back. This is a breakthrough
for a semantic layer. Not only are
we able to control the way in which measures, dimensions, attributes, and
hierarchies are displayed, but we can both allow and control the users’
abilities to write back data to the data source.
We can even store those write-backs separately from the original data,
allowing them to be viewed as deltas.
Hopefully I have succeeded in
some Pavlovian way to have you salivating at the tools available in Microsoft’s
Unified Dimensional Model. In the
next installment of this series, we will begin breaking down exactly how all of
these tools work within the system.
Go to Top |
Return to Newsletter Contents
|
