Resin transfer moulding is one of several processes available for manufacturing fibre-reinforced composites from dry fibre reinforcement. Recently, dry reinforcements made with Automated Dry Fibre Placement have been introduced into the aerospace industry. Typically, the permeability of the reinforcement is assumed to be constant throughout the dry preform geometry, whereas in reality, it possesses inevitable uncertainty due to variability in geometry. This uncertainty propagates to the uncertainty of the mould filling and the fill time, one of the important variables in resin injection. It makes characterisation of the permeability and its variability an important task for design of the resin transfer moulding process. In this study, variability of the geometry of a reinforcement manufactured using Automated Dry Fibre Placement is studied. Permeability of the manufactured preforms is measured experimentally and compared to stochastic simulations based on an analytical model and a stochastic geometry model. The simulations showed that difference between the actual geometry and the designed geometry can result in 50% reduction of the permeability. The stochastic geometry model predicts results within 20% of the experimental values.