Discrete event system simulation is often seen as a genuine tool to investigate the performance of transportation systems. The complexity of real-world systems often prevents us from accurately describing these by a mathematical model that can be evaluated analytically, thus, simulation is often the only realistic alternative. Another advantage of the simulation is the ability to include statistical analysis for different simulation scenarios.

In this paper we discuss the main problems concerning the modelling of transportation systems. Well-known approaches of incorporating uncertainty into models include trace driven simulations and sampling directly from gathered data (this latter could also be presented by a fitting statistical distribution). Another aspect to be taken into account is the economics of simulation modelling; a more detailed model requires additional building time, and proper treatment of stochastic models requires statistical analysis, which again usually implies several simulation runs. From this outset the following question arises: Should stochastic behaviour be included in transportation simulation models in the first place at all?

We present real case examples including evaluation of a railway transportation concept, capacity analysis of an automatic guided vehicle system, CBA of a railway network investment and evaluation of different multipurpose railway wagons, where stochastic behaviour is dealt with in different ways. Based on the cases we make an initial attempt to formulate framework for deciding how to include stochastic behaviour in the simulation model. We stress that the metrics used to evaluate system performance should be included in the framework. For further research topics we suggest formulating explicit guidelines to deal with stochastics to increase the efficiency of model building.