Abstract
In this paper, some of the network analysis techniques generally used for complex networks are applied to efficiency assessment. The proposed approach is units invariant and allows the computation of many interesting indexes, such as node specificity, benchmarking potential, clustering coefficient, betweenness centrality, components and layers structure, in- and out-degree distributions, etc. It also allows the visualisation of the dominance relationships within the data-set as well as the potential benchmarks and the gradual improvement paths from inefficient nodes. A number of useful filters (bipartite subgraph, ego networks, threshold networks, skeletonisation, etc.) can be applied on the network in order to highlight and focus on specific subgraphs of interest. The proposed approach provides a new perspective on efficiency analysis, one that allows not only to focus on the distance to the efficient frontier and potential targets of individual units but also to study the data-set as a whole, with its component and layer structure, its overall dominance density, etc.