This paper presents a framework for constructing regression metamodels for maintenance float systems (MFSs) with an Erlang-2 failure distribution. The metamodel is used in predicting the average equipment utilization (EU) of a maintenance float system for any given combination of repairpersons and standby units. It is shown that the regression metamodel from a 25 – 1factorial experiment yields close approximations to simulation results. TheEUmetamodel is further applied in a cost structure to determine the optimum number ofstandby units, and the number of repairpersons. Furthermore, sensitivity analysis onEUcan be conducted with the metamodel presented in this paper.