Computing gradient-based stepwise benchmarking paths

In this paper, a new stepwise benchmarking approach is presented. It is based on the concept of efficiency field potential given by a continuous and differentiable function that decreases monotonously as the amount of inputs consumed is reduced and the amount of outputs produced is increased. A gradient-based stepwise efficiency improvement method is proposed and the graphical interpretation of the continuous gradient-based trajectories is shown. A minimum potential DEA model is also formulated. The proposed approach is units invariant and can take into account preference structure, non-discretionary variables and undesirable outputs. The proposed method has been applied to an organic farming dataset.