With rapid advances in sensor and condition monitoring technologies, railways infrastructure managers are turning their attention towards the promises that digital information and big data will help them understand and manage their assets more efficiently. In addition to existing track geometry records, it is evident that track stiffness is a key physical quantity to help assess track quality and its long-term deterioration. The present paper analyses the role of the track stiffness and its spatial variability through a set of computational experiments, varying other vehicle and track physical quantities such as vehicle unsprung mass, speed and track vertical irregularities. The support stiffness conditions are obtained using a sample procedure from an Autoregressive Integrated Moving Average (ARIMA) model to generate representative larger set of data from previously on-site measured data. A set of computational experiments is carefully designed, varying different physical variables, and a vehicle-track interaction model is used to estimate track geometry deterioration rates. A series of log-linear regression models are then used to analyse the impact of the tested physical variables on the track deterioration. The main findings suggest that the spatial variability of track stiffness contributes significantly to the track deterioration rates, and thus it should be used in the future to better target design and maintenance of railway track. Finally, a comparative study of some settlement models available in literature shows that they are very dependent on the test conditions under which they have been derived.