Databases Reference
In-Depth Information
of Market Basket BI Tables is not managed, an analyst may consume all
available space with Market Basket BI Tables that are no longer needed.
For that reason, Market Basket analysts and the DBAs managing Market
Basket database must work together to avoid the junk drawer syndrome.
Analysts
The final output of the Market Basket Analysis application is query results
requested by, and delivered to, a Market Basket analyst. The analyst drives
the applications that ultimately populate the Market Basket BI Table and
consumes the data in the Market Basket Table. As such, the analyst is the
initiator and consumer of the Market Basket data.
The temptation will be strong to allow others, who are not analysts, to
view and then ultimately use the data in the Market Basket BI Table. Such
use of the Market Basket BI Table and its data should be avoided at all
costs. If non-analyst data warehouse customers require access to data in
the Market Basket BI Table, then that is a requirement for a datamart spe-
cifically for that set of customers and separate from the Market Basket
Analysis architecture. A s uccessful Market Basket Analysis application
will generate its own version of “the gravitational pull of data.” When that
happens, offer to provide the needed data in a d atamart separate from
the Market Basket Analysis architecture. If data warehouse customers are
allowed to consume the data in the Market Basket Analysis architecture,
they will eventually encumber the Market Basket Analysis application
with their requirements for data and data retention so as to render the
Market Basket Analysis unavailable for analysis.
market Basket reporting Architecture
The Reporting architectural branch, shown in Figure 6.2, also begins with a
data warehouse. Again, the data passes through a Market Basket Table and
a Market Basket BI Table. Then, any aἀ nity that varies outside the thresh-
old range is reported as an exception via an existing Exception Reporting
method. The severity associated with the variance is directly related to the
size of the variance. A small variance outside the threshold merits a lower
severity; a large variance outside the threshold merits a higher severity.
A batch job schedule, rather than an analyst, initiates Market Basket
Reporting. Market Basket Reporting focuses on the most recent set of
data available in the data warehouse. If the ETL application that loads
