The design of an industrial water system, which makes the most efficient use of the water resources available, is a complex problem that involves different trade-offs. If we assume that no fundamental process changes can be performed (i.e. wet cooling towers cannot be replaced by air-coolers, etc.), then we can improve the efficiency of the water system through practices as water re-use, regeneration of water prior to re-use, or regeneration and recycling. The inherent combinatorial nature of the problem calls for the development of a systematic methodology that can deal with the high dimensionality of the design problem. In this paper we propose an integrated methodology for the design of industrial water systems. This approach brings the engineering insights provided by the water-pinch analysis together with powerful mathematical programming tools. The method is based on a decomposition scheme for the optimisation of a superstructure model that includes all the possible features of a design. The proposed decomposition strategy is based on a recursive procedure. With this new approach, a network featuring minimum total annualised cost can be found where the complexity of the network structure is under the control of the designer and many practical constraints can be incorporated.