When determining the critical paths for the transmission of sound and vibration in assembly products, transfer path analysis (TPA) is a reliable and effective tool. TPA represents a source with a set of forces that replicate the operational responses. However, admittance-based TPA methods are prone to experimental errors, as small measurement inaccuracies can lead to large discrepancies in the source characterization. The admittance of the transfer paths is preferably obtained through impact testing. Thus, poor repeatability in the position of the successive impacts affects the consistency of the interface forces. In this study, uncontrolled location variations in a structure's excitation are characterized by a sensitivity analysis based solely on an experimental model. The functional dependency of a frequency response function on the impact location is deduced from the measured data. This makes it possible to reconstruct numerous responses for variations in the impact location and provides an appropriate sample size for the global sensitivity analysis. The influence of a random error at an individual impact location is quantified on the basis of variations in the response prediction. The approach is useful for cases where the source characterization is affected by location variations of the force input, e.g., lightly damped or complex structures where the impact locations are not easily accessed. An experimental study on an electric motor demonstrates that controlling the impact location's repeatability in a TPA is important and can lead to a more consistent source characterization.