Due to its simplicity in use and input parameters can be straightforwardly determined (from in-situ tests, laboratory tests and empirical correlations), linear elastic perfectly-plastic Mohr-Coulomb (MC) model remains a favourite constitutive modelling choice to Geotechnical engineers. Non-linear Hardening Soil (HS) model is an improvement of MC model. It is not commonly used mainly due to lack of appropriate non-linear drained stiffness parameters from laboratory tests and field calibrations. Using a deep excavation project as a case study, this paper presents a comparison of MC and HS models in assessing wall deflection. The numerical models are established based on soil parameters interpreted from available soil investigation. It was found from back analysis of a case history that laboratory tests and empirical correlations grossly underestimated the drained stiffness parameters. In order to close the gaps between predicted and observed deflections, the drained stiffness parameters have to be increased by six-fold and more than eight-fold for MC Model and HS Model, respectively.