The number of Decision Making Units (DMUs) varies widely among all the various Data
Envelopment Analysis (DEA) applications. And with increasing number of DMUs, it becomes more
and more difficult to prepare and represent the DEA results; most seriously: in case of multiple
optimal solutions. Here it is well-known that the composition of benchmarks in the so-called
envelopment-form and the weighting-scheme in the multiplier-form can be non-unique, although
the efficiency-score remains the same for the DMU under consideration. This immediately leads us
to two questions: what can be learned from these different compositions and how can we visually
In order to treat both aspects, this contribution focuses on the above stated weighting-scheme –
we call it an efficiency-configuration. From this perspective the decomposed objective function
value helps us to localize the DMU’s activity specialization. A technique called Multidimensional
Scaling then allows for a visual identification of clusters among all DMUs with respect to their
specialization – more precisely, their efficiency-configuration. But such a clustering should be
studied in the presence of multiple optimal solutions. This is topic of the present contribution.
Finally, after developing the respective theoretical concepts, a real-world application will be
examined and discussed.
Keywords: efficiency-configuration -- multidimensional scaling -- clustering -- visualization