This paper discusses the development of a methodology to generate drive cycles having a finite duration, but which are statistically representative of a larger set of usage data collected from fleet vehicles operating in the field. Given field-generated time vs. velocity data, acceleration at each data point is calculated, and each velocity and acceleration pair is binned using some calibrated level of fidelity. As a result, a velocity-acceleration matrix representing each vehicle operating point, as well as cumulative probability distribution functions for acceleration change and take-off acceleration are generated. These cumulative distribution functions are utilized to pick random velocity-acceleration pairs from the corresponding matrix, and the concatenation of each consecutive chosen velocity-acceleration pair constitutes the final drive cycle. Three drive cycles representing the high-, medium- and low-speed operation of the vehicle are generated from the field data, and these show measurable similar statistical characteristics to the initial master data superset. To validate the drive cycles, simulations were performed on a representative vehicle model using both the statistically generated drive cycles and the entirety of the field-collected dataset, and fuel economy results were found to be consistent between the two. Finally, a method for evaluating various use cases of the vehicle in terms of fuel economy using various combinations of these three drive cycles is discussed.