Databases Reference
In-Depth Information
the presence of the gravitational pull of its own data. In that way, a data
warehouse purpose can become its own self-fulfilling prophecy.
Build it and they will come. Yes, that often works, but to what level of
success? Each subject area of a data warehouse must have its own purpose.
Without a purpose to act as cohesive glue, a subject area of a data warehouse
will eventually begin to spin out of control. After only a few subject areas
experience this form of orbital decay, a data warehouse is already at risk.
he purpose of this topic is to suggest and explain two purposes that
can increase the value and ROI of a data warehouse. Specifically, those two
purposes are Market Basket Analysis and Time Variant Data. Many enter-
prises aspire to include Market Basket Analysis in their set of analytics.
Some enterprises pay marketing and consulting firms to provide Market
Basket Analysis. Unfortunately, the results delivered often fail to meet the
expectations, and sometimes fail to deliver at all. The following chapters
will explain the obstacles to Market Basket Analysis and a design solution
that overcomes those obstacles.
Time Variance is one of the concepts included in the pioneering works
of Inmon and Kimball. The idea is that a data warehouse will present his-
torical data in its historical context. For example, transactions in 1995
will be presented in their 1995 context while transactions in 2011 will be
presented in their 2011 context. While the concept of Time Variant Data
is intuitively easy to understand, it is also diá¼€ cult to deliver each and
every transaction in its historical context. An individual query may span
a period of weeks, months, or years, which would require it to include the
Time Variant context for each of those transactions. The solution design
for Time Variant Data overcomes the obstacles of Time Variance, allowing
a data warehouse to deliver Time Variant Data with a nominal degrada-
tion in performance.
Finally, the last chapter will combine these two purposes into one deliv-
ery. As stated earlier, when you hit the bull's-eye in the center, there is
more bull's-eye in every direction. The combination of Market Basket
Analysis using Time Variant Data is an example of such an ancillary
benefit. By delivering the purposes of Market Basket Analysis and Time
Variant Data, a data warehouse can achieve another deliverable—Market
Basket Analysis of Time Variant Data.
A data warehouse that includes those three deliverables (Market Basket
Analysis, Time Variant Data, and Market Basket Analysis of Time Variant
Data) is a data warehouse with purpose. A data warehouse that delivers
those three purposes is a data warehouse with an ROI.
