This contribution analyses the state of the art in two parts. Based on general quasi-homogeneous and heterogeneous reactor models, their benefits and disadvantages as well as simplification options are discussed, taking into consideration catalyst deactivation. Further on, the article focuses on the issue of model trimming for industrial applications. A distinction is made between process rationalisation and the development of new processes. In process rationalisation, models can be tested/adjusted on the basis of measured values from industrial reactors. This is not feasible when developing new processes. Thus, the major factors influencing the reliability of prediction, such as model assumptions, initial and boundary conditions, mathematical and computation restrictions, chemical kinetics, thermal and mass transport as well as flow distribution, are analysed first. Eventually, this contribution proposes ways of model trimming and resolving problems of scale transfer.