An acceleration- based procedure for accurate determination of propulsion system performance from flight daja is described in this paper. The procedure emplpys iteration techniques to achieve a weighted least-squares ("best") fit to all of the applicable flight data. In the application of the procedure to the USAF Titan III launch vehicle flight programs, several analysis problems were encountered including ^termination of propellant level sensor effective uncover time, ablative engine throat variations, instrumentation drift and bias errors, inaccurate subsystem mathematical models, and determination of outage and cutoff impulse for propellant depletion shutdowns. Methods of treating these problems are described which permit determination of performance-related results with high accuracy, often from poor-quality or substitute data. In contrast to pressure-based procedures which frequently generate errors exceeding 2%, the over-all process described usually achieves an accuracy within J% in the determination 'of thrust, specific impulse, and mixture ratio. These techniques which have evolved during the flight programs have permitted refinement of Titan III subsystem models, yielding significant improvements in vehicle flight prediction accuracies.