The selection of a suitable manufacturing process often involves considering the complex coupling between characteristics of the design, the material and the process. Whilst most materials can be well described by a common set of properties, enabling selection for a given design on the basis of these properties alone, the same is only partially true for process selection. The most discriminating characteristics of processes are often specific to the class of process. For example, very different questions arise when selecting a casting process than when selecting a welding process, so the information needed to answer these questions is mostly specific to each process class. Furthermore, the data and information needed to capture these characteristics can be strongly influenced by the class of material being processed there is limited scope for selecting a welding process for aluminium, or steel, or polymers from a generic welding selector that does not have material-specific data. This paper considers the general problem of building selection tools for specific manufacturing "tasks". A task is defined as a subset of processes applied to a subset of materials. The goal is to identify systematically the match between the requirements of the design and the capabilities of processes. A methodology has been proposed for this task-based process selection which involves consideration of the attributes of the material, design and process which are relevant to the task in hand. Three levels of quantitative requirement-attribute coupling are identified for selection at the task-level in design. Coupling involving only two or three attributes can be handled by construction of suitable task-specific process databases. More complex interactions require a different approach, in which modelling plays a key role in capturing the relationships between the design features, the material behaviour during and after processing, and the process parameters. Modelling is interpreted here in its widest sense: from empirical rules and curve fits, to advanced statistical methods such as neural networks, to physically-based process models. The use of modelling opens up great opportunities for making maximum use of sparse process data, for optimum co-selection of material and process, and for providing the designer with feedback on the likely influence of processing on the viability and cost of a design as well as indicating trial processing parameters. This review discusses the general issues with reference to a range of previously studied selection problems including aluminium casting, joining, welding, and heat treatment of steels. The role of modelling in enhancing selection is illustrated for cutting and welding of carbon steer, and future potential developments are discussed.