[Back]

J. Pichler:
"Multiobjective-Driven Decision Support for Selection of Business Application Portfolios";
Supervisor: A. Tjoa, T. Neubauer; Institut für Softwaretechnik und Interaktive Systeme, 2007; final examination: 2007-04-20.

Because in the last decade, a strong coherence between business and IT as well
as a thorough commitment to IT have become an important factor of competition
on all market places and in nearly all industries, companies spent enormous
amounts of money in information technology infrastructure [33]. Because of this,
in the last few years many companies started to question their IT´s real value
and the total of benefits it generates. This usually induces a problem since modern
software exposes many intangible characteristics that cannot be measured
on common scales (such as usability, reliability or security). By that, valuation
of software frequently becomes a complex, in some cases intractable tasks.
Many modern IT management and aligning methodologies (such as IT portfolio
management) emphasise a holistic management of IT landscape, which of course
includes valuation of all employed software.
Today, there are multiple structured software selection approaches that enable
evaluation of software. Though most of these methods - such as the Analytical
Hierarchic Process (AHP) or Weighted Scoring Method (WSM) - permit finding
the "best" out of a certain amount of candidates, they suffer from major
disadvantages which render them unusable for repeated and sustained valuation
of corporate software portfolios. Current methods for software evaluation employ
the technique of preference weighting thus requiring an a-priori definition of
objective rankings. This imposes the problem of decision makers having to formulate
all restrictions prior to calculating candidate weights, which can result in
the algorithms´ having a low repeatability and their outputs´ being biased to an
uncertain extent. In addition to this, quality of solutions directly depends on the
quality of a-priori data which makes the process of input data collection extremely
sensitive. Another major drawback of current algorithms is that they are limited
to evaluation of one single candidate at a time: Because most companies employ
multiple software packages simultaneously, algorithms are demanded to consider
groups (portfolios) of candidates together with their mutual dependencies.
Utilising the paradigms of portfolio theory and multiobjective optimisation,
this thesis aims at developing a powerful yet straight-forward portfolio-aware
software selection model that permits automatic, software-based evaluation of
all possible candidate combinations under alignment to both corporate-level and
project-level constraints. Multiobjective approaches do not aggregate criteria of
different types into over-all values, nor do they require a-priori induction of preferences
(e.g. objective weights). Thus, the new model respects the candidates´ real objective values and permits dynamic definition of preferences a-posteriori.
Because no extensive collection of a-priori data is necessary, the approach yields
high reusability permitting evaluation of whole corporate software portfolios on
a regular basis with little effort and high transparency. This significantly improves
manageability of corporate IT, reducing costs and leveraging efficiency of
corporate business.
To test the new model´s functionality, a decision support application is implemented
to solve a real-world decision situation. Results are compared to those of
two popular software valuation approaches, AHP and WSM.