Traditional modelling notions presume the existence of a ‘truth’ model that relates the input to the output, without advanced knowledge of the input. This has led to the evolution of education and research approaches (including the available control and robustness theories) that treat the modelling and control design as separate problems. This paper explores the subtleties of this presumption. A detailed study of the nature of the modelling errors is useful to gain insight into the limitations of traditional control and identification points of view. Modelling errors need not be ‘small’ but simply ‘appropriate’ for control design. Furthermore, the modelling and control design processes are inevitably iterative in nature.