Liquid Composite Moulding (LCM) process design requires that the resin impregnation minimizes both mould filling time, as well as the probability of occurrence of dry, unsaturated zones. This can be challenging, as the degree of variability present in material properties and the manufacturing process itself, can be substantial. The main purpose of this study is to understand and quantify how the different sources of variability, present in LCM, may affect the performance of the mould filling process, creating defects and to quantify how lower scale interactions may affect macro-scale flow. In order to transfer the meso-scale stochastic behaviour towards a complete macro-scale model, an uncertainty propagation model is proposed, whereby local statistical properties are used to predict the upscaled flow behaviour. This is done by means of flow simulation, using stochastic models for the input process variables. Local fabric distortions, combined with fabric permeability, pressure and race-tracking serving as inputs to estimate the joint probability distribution that characterizes the global uncertainty in flow performance. Preliminary uncertainty analyses investigating individual sources of variability are available in the literature, however, a complete framework combining the different variables into a mould filling scenario, was not found. Therefore, this study tries to develop a new multi-variable aggregating concept that is applied to mould filling simulation. The individual contribution of each source of variability used in this work is quantified, providing a better understanding of the reliability of alternative process designs.