We analyze the technical efficiency of German and Swiss urban public transport companies by means of SFA. In transport networks we might face different network structures or complexities, not observed, but influencing the production process. The unobserved factors are typically modeled as separable factors. However, we argue that the entire production process is organized around different network structures. Therefore, they are inevitably non-separable from the observed inputs and outputs. The adopted econometric model is a random coefficient stochastic frontier model. We estimate an input distance function for the years 1991 to 2006. The results underline the presence of unobserved non-separable factors.