My first Publication Agile-Data-Warehouse-Design-eBook | Page 40

How to Model a Data Warehouse
19
Agile Dimensional Modeling and Traditional DW/BI Analysis
Agile dimensional modeling doesn’t completely replace traditional DW/BI analysis
tasks, but by preceding both data-driven and reporting-driven analysis it can make
them agile too: significantly reducing the work involved while improving the
quality and value of the results.
Agile Data-Driven Analysis
Agile data-driven analysis is streamlined by targeted data profiling. Only the data
sources implicated by the agile data model need to be analyzed within each itera-
tion. This targeted profiling supports the agile practice of test-driven development
(TDD) by identifying the data sources that will be used to test the data warehouse
design and ETL processes ahead of any detailed physical data modeling. If an ETL
test can’t be defined because a source isn’t viable, agile data modelers don’t waste
time physically modeling what can’t be tested, unless they are doing proactive data
warehouse design. In this case the agile data warehouse model can assist the test-
driven development of the new OLTP system.
Agile Reporting-Driven Analysis
Agile reporting-driven analysis takes the form of BI prototyping. The early delivery
of dimensional database schemas enables the early extraction, transformations and
loading (ETL) of real sample data so that better report requirements can be proto-
typed using the BI user’s actual BI toolset rather than mocked-up with spread-
sheets or word processors. It is intrinsically fairer to ask users to define their
requirements and developers to commit to them, once everyone has a sense of
what their BI tools are capable of, given the available data.
Agile dimensional
modeling makes
traditional analysis
tasks agile
Data-driven analysis
becomes targeted
data profiling
Reporting-driven
analysis becomes
BI prototyping
Requirements for Agile Dimensional Modeling
Agile modeling requires both IT and business stakeholders to change their work
practices and adopt new tools and techniques:
Collaborative data modeling requires open-minded people. Data modelers
must be prepared to meet regularly with stakeholders (take on a business ana-
lyst role) while business analysts and stakeholders must be willing to actively
participate in some data modeling too. Everyone involved needs simple frame-
works, checklists and guidelines that encourage interaction and prompt them
through unfamiliar territory. Collaborative
Business stakeholders have little appetite for traditional data models, even
conceptual models (see Data Model Types , shortly) that are supposedly targeted
at them. They find the ER diagrams and notation favored by data modelers
(and generated by database modeling tools) too complex or too abstract. To
engage stakeholders, agile modelers need to create less abstract, more inclusive
data models using simple tools that are easy to use, and easy to share. These in-
clusive models must easily translate into the more technically detailed, Collaborative
modelers require
techniques that
encourage
interaction
data modeling
must use simple,
inclusive notation
and tools