Multiunit restaurant productivity assessment using three-phase data envelopment analysis

This paper focuses on uncontrollable variables’ effects on multiunit restaurant productivity using data envelopment analysis (DEA). We argue the importance of first considering managerially uncontrollable (nondiscretionary) variables as inputs in the actual DEA model, with managerially controllable variables considered post hoc for their relationship to the efficiency scores. We illustrate the merits of this approach using data from a chain of 62 full-service restaurants. From a large number of candidate inputs, we arrive at a short list of uncontrollable inputs: hourly server wage, restaurant seats, and a coding variable representing whether the restaurant is a stand-alone facility. Output variables in our model were daily sales and tip percentage. We find that just under 12% of the restaurants operate efficiently and that the average efficiency for the chain is 82%.