Database Reference
In-Depth Information
eXPlAnATIon oF TeRMS
The field of ad hoc analytics is vast and murky, with different vendors per-
petuating their own terminology. Microsoft, as always, does the same, and it
is important to understand the different way the tools are defined.
First, ad hoc analytics needs to be differentiated from the way the term analyt-
ics is used in general. The term is often used as a synonym for predictive ana-
lytics, statistical analysis, or machine learning. These fields (to some extent one
and the same) are all based around the concept that an analysis is being done
mathematically and probabilistically—far outside the scope of this chapter.
Instead,
ad hoc analytics
is the analysis by a person of a data set, by dynami-
cally applying filters, grouping, and aggregations until a result is discovered.
This process, as mentioned, is also called slice and dice, and (outside the
Microsoft space) is called data mining as well.
Data mining
in that context is
analyzing data to derive value. However, Microsoft has a tool as part of its
Analysis Services called Data Mining Extensions and a language DMX, which
are in fact a statistical toolset. This distinction is important as it can lead to
some confusion!
Some other important terms that are often confused are covered next.
SELF-SErVICE BI
Self-service BI is another one of those hot industry topics with confusion
attached.
In the first definition, self-service BI simply refers to the ad hoc analytic capa-
bility described in the preceding section. Data has been prepared, possibly in
a data warehouse, loaded into a BI tool, perhaps OLAP or not, and then the
business users are allowed to interact with the data in this controlled manner.
At the other end of the spectrum is the self-service user who is gathering his
own data and then building his own reports. Typically, as seen in Figure 13-1,
this is easy enough when a single source of data (such as an ERP system) is
available, but this becomes much more difficult after multiple sources of data
are required to be combined, such as a CRM and an ERP system, with the pos-
sibility of duplication and data mismatches.
