In the spring of 1989, the Air Force contracted for the development of a suite of high fidelity maintenance trainers to support the training and certification of maintenance technicians for the C-17 air-lifter. This set of 11 trainers encompassed the simulation and replication of every subsystem of the still-in-development C-17 aircraft. The development of the C-17 Maintenance Training Devices (MTD) would be the first attempt on such a large scale to support certification of technicians on equipment other than the aircraft. Tenets of the program included: minimizing the use of aircraft parts, causing a great dependence on aircraft data; no formal training on the aircraft, thus eliminating parsing of tasks between aircraft and trainer; and delivery of the trainers concurrent with deployment of the aircraft to Charleston AFB, SC. As the amount of training time available on modern aircraft decreases, and formal training programs extend to the flight line, training/certification devices with comparable fidelity requirements will proliferate (i.e., MV-22, F/A-18E/F, F-22). This paper discusses some lessons learned from the development and test of the C-17 MTD program, with emphasis on the definition, design, and test of adequate fidelity, to support the certification requirements of the user (specifically, Air Mobility Command). In retrospect, a recurring theme throughout the various phases of the C-17 MTD development, is that the evaluation of higher level requirements and the definition of lower level requirements continues on through the test program. Hopefully, this paper will evoke some thought for harnessing the inevitability of this requirements process so that future programs will result in products that meet the user's requirements and expectations.