Road slips are common in Northland Allochthon residual clay soil, and are commonly mitigated using deep soil mixing (DSM). A deficiency in laboratory investigations on Northland Allochthon residual clay and a need for a better understanding of the numerical modelling of DSM columns used to mitigate unstable slopes in this soil type is evident in literature, and has been highlighted by practitioners. This research has aimed to fill aspects of these deficiencies. Field testing and classification tests have provided insight into how the soil varies between sites and with depth, and how in situ testing methods compare to one another. Field testing has also demonstrated that soil property changes around DSM columns have been shown to exist through seismic flat plate dilatometer testing before and after column installation, which has not previously been proven using an in situ method. This is important for practitioners who use DSM to demonstrate the additional soil improvements provided by the columns. The testing of reconstituted soil is fundamental in examining soil behaviour, and this study is the first to examine the triaxial behaviour of reconstituted specimens of Northland Allochthon soil. Laboratory triaxial testing and oedometer testing have allowed for a normalized comparison of the intact strength of Northland Allochthon residual clay soil to its reconstituted state. This work provides an answer to the important question regarding the role of soil structure in this soil type. It was revealed that soil structure results in increased shear strength of the soil, and that this increase is primarily cohesive in nature. The near coincidence of the post-rupture strength of intact specimens with the critical state angle of internal shearing resistance provides support for its use in examining first time slope failures in this soil type. This is an important finding for practitioners, as it demonstrates the value of testing reconstituted specimens, which are much easier to obtain than high quality intact specimens. In addition, relationships between the plasticity index (PI) of the soil and certain soil parameters (and soil behaviour) have been demonstrated to be relevant and useful for this soil type. Soil properties acquired in this study were tabulated along with those from other field sites in Northland Allochthon soil. It was found that there is significant variation between field sites, likely due to varying degrees of weathering, which is an important consideration for practitioners dealing with this soil type. A brief examination of constitutive models for representation of Northland Allochthon residual clay soil have shown that several different models can sufficiently represent the behaviour of this soil. The Mohr-Coulomb model was selected for use in subsequent finite element numerical models. A case study of a road slip at a field site in Northland Allochthon residual clay soil, mitigated using DSM columns, revealed that the use of a pre-existing slip surface after first time failure leads to an improved match between observed field behaviour and the behaviour of the slope as exhibited in a numerical model. This type of failure mechanism has not been previously examined in this soil type, and this case study demonstrates it is a useful approach that should be considered when dealing with second time failure in Northland Allochthon slopes. This numerical model also introduces the replacement ratio method (RRM), a technique used to represent the three dimensional (3D) geometry of the DSM columns in the more commonly used two dimensional (2D) analysis. Examination of laterally loaded DSM columns in plan view, which has not previously been performed in the context of DSM columns, has illustrated how installation effects and column shape influence load displacement curves, and demonstrates the effects of soil arching. This analysis provides practitioners with evidence that improved soil property changes, found to occur around DSM columns, lead to improved DSM column performance. A simplified 3D numerical model of laterally loaded DSM columns, which builds on the ideas developed in the previous two 2D models, has been compared to an identical 2D model. It is shown that the commonly used RRM results in an overestimation of the resisting force provided by the columns as compared to the 3D model. However, this does not necessarily imply that the use of the RRM in an analysis will always result in a safe slope. The degree to which its use will affect the results will depend on the slope geometry, location of the DSM columns, and the type of analysis performed (i.e. factor of safety or deformation based). A modification to the RRM has been proposed. It is recommended that when the DSM column diameter and soil properties are similar to those used in this study, the MRRM developed in this study should be utilized. In circumstances where they differ, it is recommended that practitioners perform a sensitivity analysis using the MRRM developed here as a basis for modifying the RRM in order to determine the extent to which their results are influenced. If the influence is significant, the use of a 3D model should be considered.