My first Publication Agile-Data-Warehouse-Design-eBook | Page 118

Modeling Business Processes
97
development. But the problem is this fallback to the “big design upfront” (BDUF)
simply does not match the evolutionary nature of modern BI requirements nor
their delivery timescales. Plus it is incredibly hard for a DW/BI project to become
agile when it does not start off agile.
Instead, agile data warehouse modelers should stay agile (and dimensional), but
lower their technical debt by balancing “just in time” (JIT) detailed modeling of
business events for the next development sprint and “just enough design up front”
(JEDUF) for cross-process BI in the future. To do so, modelers need to rapidly
model ahead in just enough detail to discover which of the dimensions, needed for
the next sprint, should also be conformed dimensions that will help to future proof
their designs for enterprise BI.
Agile dimensional
modelers lower their
technical debt by
modeling ahead just
enough to define
conformed
dimensions
Conformed Dimensions
Figure 4-2 shows a Promotion Analysis Report that combines information from two
events: CUSTOMER ORDERS and PRODUCT CAMPAIGNS to explore the
connection between campaign activity and sales revenue. The report is possible
because the two different events have identical descriptions of PRODUCT and
PROMOTION. These conformed dimensions allow measures from both events to
be aggregated to a compatible level and lined up next to one another on the report.
Lining up the answers or drilling-across like this appears obvious but if the events
are handled by different operational systems (an Oracle-based order processing
application and a SQL Server-based customer relationship management system)
then this report might be the first time that the two sets of data have actually met.
If each source system describes products and promotions differently and the
individual star schemas use these non-conformed descriptions, the analysis would
not be possible because the measures would not align.
Conformed
dimensions allow
measures from
different events to
be combined and
compared
Figure 4-2
Conformed
dimensions
enable
cross-process
analysis