Sustainable urban water supply management requires, ideally, accurate evidence based estimations on per capita consumption and a good understanding of the factors influencing the consumption. The information can then be used to achieve improved water demand forecasts. Water consumption patterns in the developed countries have been extensively investigated. However, very little is known for the developing world. This paper investigates per capita water consumption resulting from water use activities in different types of households typically found in urban areas of the developing world. A data collection programme was executed for 407 households to extract information on household characteristics, water user behaviour and intensity and the nature of indoor and outdoor water use activities. The rigorous statistical analysis of the data shows that per capita water consumption increases with income: 241, 272 and 290 l/capita/day for low, medium and high income households, respectively. Additionally, the results suggest that per capita consumption increases with the number of adult female members in the household and almost one-third of consumption is via taps. The collected data has been used to develop statistical models using two different regression techniques: multiple linear (STEPWISE) and evolutionary polynomial regression (EPR). The inclusion of demographic parameters in the developed models considerably improved the prediction accuracy. Two of the best performing models are used to forecast the water demand for the city, using four future scenarios: market forces, fortress world, policy reform and great transition. The results suggest that the domestic water demand would be highest in the fortress world scenario due to the increase in population and size of built-up area.