Industrial steel storage racks are pre-engineered lightweight structures commonly used to store goods from supermarkets to big warehouses. These systems are framed structures, usually made of cold-formed steel profiles and characterised by non-standard details. Their performance is quite complex and the prediction of their global response is more difficult than for the traditional steel frames. This difficulty is due to the racks’ main features: the use of cold-formed thin-walled steel sections which are sensitive to different buckling modes, the presence of regular perforation patterns on the uprights, the highly non-linear behaviour of joints, the influence of the structural imperfections and the significant frame sensitivity to second order effects. The behaviour of racks becomes even more complex when seismic or accidental events induce significant horizontal forces acting on the structures. The complexity and variability that characterise racks make it difficult to identify general design solutions. Hence, racks design is traditionally carried out by using the “design by testing” approach, which requires the experimental characterisation of the main structural components, of the joints and the sub-assemblies. The complexity of the racks also affects their numerical modelling, which results in complex analyses that must take into account all the aforementioned features. The work presented in this thesis focuses on the study of a typical steel pallet rack, identified as case study. The research aims to contribute to building up a comprehensive knowledge of the response of both the main rack components and of the whole structure. The main rack components were first individually studied. The behaviour of the uprights, of the base-plate joints and of the beam-to-column joints was experimentally investigated. The experimental data were then taken as reference for the calibration of FE models that enabled exploring each component’s performance. These models were then incorporated into the whole rack model. The response of the uprights was first investigated through stub column tests. The non-negligible interaction between axial force and bending moment of the upright response was then experimentally and numerically analysed to define the M-N domains. In addition, the rules provided by different European standards for the design of isolated members subjected to combined axial load and bending moment were considered and critically compared, identifying the main critical issues of the different design approaches. Although the contribution of joints on the rack global response is of paramount importance, to date, the knowledge is quite limited. In particular, the experimental studies of the behaviour of base-plate joints are still rather modest, especially for the cyclic range. Therefore, an experimental campaign on the rack base-plate joints was carried out: three levels of axial load were considered and the response in both the down-aisle and the cross-aisle direction was investigated under monotonic and cyclic loadings. Similarly, the beam-to-column joint was tested both monotonically and cyclically, taking into account its non-symmetric behaviour. Numerical models for both joint types were developed and validated enabling the characterisation of joints in the monotonic and cyclic range. This in-depth knowledge of the response of individual components facilitated the evaluation of the global rack behaviour. As a final stage of the research, full-scale tests of four-level two-bay racks were performed taking advantage of an innovative full-scale testing set-up and, on the basis of the experimental outcomes, the racks’ global behaviour was numerically investigated. Critical standards issues and needs for future research were further identified.