Data envelopment analysis (DEA) is a system-based approach to performance assessment that considers multiple inputs and outputs simultaneously. In spite the strength and a history of 25 years of theoretical foundation, most practitioners still utilize the output-input ratio (partial productivity) for assessment. The major hurdle in making this switch is the required paradigm shift - theoretical understanding of DEA requires economics and mathematical programming background, and the results are less intuitive than traditional partial productivity measures. To fuse the two approaches and simplify their joint use for practitioners, the relationship between efficiency scores provided by DEA, which corresponds to the economic concept of technical efficiency (TE), and conventional partial efficiency (PE) must be explained. Potential information loss in both methods are compared and contrasted.

The main objective of this dissertation is to accomplish the fusion by answering the question "What is the relationship between the efficiency scores provided by DEA and the partial productivity metrics?" We first connect TE and PE assuming a priori costs/prices information. We show PE is a special case of bilateral comparison of the performance of two organizations. Integrating the early works, the bilateral comparison, thus, can be further decomposed into detail comparisons based on eleven metrics. We also build the connection between TE and PE directly. We show that both PE and TE can be computed using similar LP formulation. Therefore, a sequence of LP models, which starts from TE and ends at PE, collectively, forms a bridge between PE and TE. This bridge has several "spans", each corresponding to a particular effect. Therefore, a particular PE can be decomposed into seven multiplicative factors including TE. This theoretical linkage provides aids for output-input ratio benchmarking performance gap analysis and leads to some practical guidelines for selecting a PE to approximate system-based efficiency when DEA is not a possible solution. In the end, we present a warehouse benchmarking study.