Due to the availability of commercial finite element and finite difference codes which are, due to high level graphical user interfaces, easy to handle, numerical methods are routinely applied in geotechnical engineering to assess the load displacement behaviour of geotechnical structures, e.g. excavations, tunnels and foundations. However, simple failure criteria such as the Mohr-Coulomb criterion are not sufficient for representing the complex stress strain behaviour of soils for stress states well below failure and thus advanced constitutive models are increasingly employed in practice. This leads to the problem of parameter identification because advanced models need more, and sometimes different, parameters than usually available in a standard geotechnical report, which in general do not reflect these developments in numerical modelling. This leaves the designer with the difficult task of determining input parameters for numerical analyses based on, at least in many cases, limited information. Some practitioners therefore favour the development of a public data base where parameters for common material models for different soils could be made available, but to the author's opinion this is not the way forward. Another option discussed is the extended use of correlations based on in-situ testing.