This paper uses a sensitivity analysis to quantify the dominant train properties (mass and spacing of wheels and bogies) that contribute to ground-borne vibration generation, with the aim of reducing the complexity of train–track numerical models. This research is significant because ground-borne vibration from railways is a growing problem, particularly in urban areas. Despite this fact, attempting to predict vibration levels is complex because there are many variables that contribute to the overall dynamic response. Therefore, a deterministic approach is commonly used, that ignores many of these variables. Thus, this paper identifies the variables that can be ignored, while highlighting those that are highly influential on vibration generation. For this purpose, a previously validated 2.5D finite elements-boundary elements approach is used to simulate dynamic train–track interaction. It is computed many times for a variety of modelling variables to investigate the effect of each on the ground-borne vibration levels in the far field. It is found that increases in unsprung mass of the train causes a large increase in vibration levels. Furthermore, changes in wheel/bogie spacing and semi-sprung mass are found to have a minimal effect on vibration generation.