Contemporary geotechnical design often requires the use of advanced numerical analysis, if it is to take account of the complex nature of many geotechnical problems. One crucial aspect of such analyses is the realistic representation of the facets of soil behaviour that are dominant in any given problem, which in turn requires a careful selection of an appropriate constitutive model and derivation of model parameters from the available, and often disparate, experimental data. This paper uses the authors’ experience of advanced numerical analysis and constitutive modelling to emphasise the importance of close integration of the process involved with interpreting experimental data with the process of selecting and calibrating advanced constitutive models, in successfully predicting the response of geotechnical structures.