My first Publication Agile-Data-Warehouse-Design-eBook | Page 32

How to Model a Data Warehouse
11
Data Warehouse Analysis and Design
Both 3NF ER modeling and dimensional modeling are primarily database design
techniques (one arguably more suited to data warehouse design than the other).
Prior to using either to design data structures for meeting BI information require-
ments, some form of analysis is required to discover these requirements. The two
approaches commonly used to obtain data warehousing requirements are data-
driven analysis (also known as supply driven) and reporting-driven analysis (also
known as demand driven). While most modern data warehousing initiatives use
some combination of the two, Figure 1-4 shows the analysis and design bias of
early 3NF enterprise data warehouses compared to that of more recent dimen-
sional data warehouses and data marts.
Analysis
techniques are
required to
discover BI data
requirements
Figure 1-4
Data warehouse
analysis and design
biases
Data-Driven Analysis
Using a data-driven approach, data requirements are obtained by analyzing oper-
ational data sources. This form of analysis was adopted by many early IT-lead data
warehousing initiatives to the exclusion of all others. User involvement was
avoided as it was mistakenly felt that data warehouse design was simply a matter of
re-modeling multiple data sources using ER techniques to produce a single ‘per-
fect’ 3NF model. Only after that was built, would it then be time to approach the
users for their BI requirements.
Pure data-driven
analysis avoided
early user
involvement