A learning ladder toward efficiency: Proposing network-based stepwise benchmark selection

• We propose an application of Social Network Analysis (SNA) in the DEA context.
• The proposal, grounded on the knowledge-based view, considers efficiency improvement as a learning process.
• The method highlights the possible existence of outliers.

Starting from the knowledge-based view of efficiency improvement, we propose a network-based approach to find the optimal stepwise benchmarking paths toward the efficiency frontier. The approach treats the Data Envelopment Analysis system as a network of teaching and learning firms and calculates the overall shortest paths taking into account both input endowment similarity and the efficiency gap covered in each step. In addition, based on network centrality concepts, the method discriminates between efficient and intermediate units, and highlights possible outliers or specialized units. As a real-world example, the method is applied to a network of Canadian bank branches and practical implications are discussed.