From User Requirements to Conceptual Design in Data Warehouse Design - Data Warehousing Design and Advanced Engineering Applications - page 6

compound KPIs is more common. On the other
hand, the supply-driven approach is simpler and
cheaper (in terms of both time and money) than
other approaches since its duration only depends
on the designer skills and on the data sources
complexity. A further strength of this approach
is the quality of the resulting multidimensional
model that will be very stable since it is based
on the schema of the operational data sources
that do not change as frequently as the personal
requirements of the business users.
The user-driven approach is a bottom-up
technique that starts from determining the in-
formation requirements of different business
users (Winter & Strauch, 2003). Their points of
view are then integrated and made consistent in
order to obtain a unique set of multidimensional
schemata. The emphasis is on the requirement
analysis process and on the approaches for fa-
cilitating user participations. The problem of
mapping these requirements onto the available
data sources is faced only a posteriori, and may
fail, thus determining the users' disappointment.
Although this approach is highly appreciated by
business users that feel involved in the design and
can understand the rationale of the choices, it is
usually time expansive since the business users
at the tactical level rarely have a clear and shared
understanding of the business goals, processes and
organization. Consequently, this approach usually
requires great effort by the project manager, that
must have very good moderation and leadership
skills, in order to integrate the different points of
view. Furthermore, the risk of obsolescence of
the resulting schemata is high if requirements are
based on the personal points of view of the users
and do not express the company culture and the
working procedures.
The g oal-driven approach focuses on the busi-
ness strategy that is extrapolated by interviewing
the top-management. Different visions are then
analyzed and merged in order to obtain a consis-
tent picture and finally translated into quantifiable
KPIs. This approach (Boehnlein & Ulbrich vom

Ende, 2000) is typically top-down since by starting
from the analysis of a few key business processes,
their characterizing measurements are derived
first and than transformed into a data model that
includes a wider set of KPIs that characterize such
processes at all the organizational levels. The ap-
plicability of this approach strictly depends on the
willingness of the top management to participate to
the design process and usually require the capabil-
ity of the project staff in translating the collected
high-level requirements into quantifiable KPIs.
Goal-oriented approaches maximize the prob-
ability of a correct identification of the relevant
indicators, thus reducing the risk of obsolescence
of the multidimensional schema.
In many real cases the difference between
adopting a goal-driven instead of a user-driven ap-
proach may become very vague, on the other hand,
it should be clear that the goal-driven process is
top-down and based on the progressive refinement
of a few goals defined by the top-managements,
while in the user-driven approach, requirements
are obtained by merging several simpler require-
ments gathered from the business-users in a
bottom-up fashion. The result of a goal-driven
approach differs from a user-driven one whenever
the users do not have a clear understanding of the
business strategy and the organization's goals.
Table 1 reports a comparison of the three basic
approaches and may be useful to choose the one
that is most suited to a given project. The main
technical element influencing such a choice con-
cerns the availability and the quality of the schema
of the operational data sources, while several
non-technical factors are involved in the choice.
In particular, the cost and time constraints sug-
gest a reduction of the time devoted to interviews
and discussions with the users; similarly when
the business users have a limited knowledge of
the business process and strategy a user-driven
approach should be avoided.
In order to avoid the drawbacks of the single
approaches some mixed strategies have been de-
veloped. In particular, Bonifati et al. (2001) mix the