The paper describes a statistical approach and the developed software tool for the cost evaluation of alternative distribution network design strategies. In contrast to traditional approaches to optimal network design, which are based either on the analysis of a small specific service area or on idealistic networks, the proposed method determines optimal network design polices by evaluating alternative strategies on many statistically similar networks. The software can generate a number of realistic consumers sets that are characteristic for urban and rural areas, in terms of consumer distributions, types, numbers etc. These sets are then supplied by a large number of different networks following several alternative design strategies. Evaluating the cost of each particular design over a number of statistically similar networks allows statistically significant conclusions to be drawn. Simulations have been carried out on a single voltage level and on two voltage levels, considering different load densities and the use of underground cable and indoor substations or overhead line and outdoor substations. The results have been analysed to give an indication of the optimum electricity distribution design policy in terms of the cost of equipment, its installation and maintenance, and variable and fixed losses. The software is currently being used at one of the UK electricity companies to review distribution network design.