Transforming smart meter data captured at low-resolution litre intervals of 15 to 60 mins into residential water end use data provides valuable insights for water businesses and their customers. Water end use data is crucial for developing effective and customised water conservation and management strategies. In this study, a significant water end use event data repository collated from studies covering various Australian metropolitan cities was used to develop an intelligent model for low-resolution water consumption data based on Volume and Time features. The model architecture integrates the strengths of Random Forest, Linear Regression, and Neural Networks in a stacked ensemble, with a Regression Tree as a meta-model. Model accuracy was found to stabilise with of sufficient size and observation period. The model was then applied to a larger dataset, enabling a robust case study. The results demonstrate that the model accurately predicts end use categories, offering valuable insights into large-scale water consumption behaviours. This research underscores the importance of feature selection and optimal dataset size in enhancing model accuracy. The derived residential water end use model makes a significant contribution by allowing water businesses to autonomously and accurately characterise citywide residential end uses using only mainstream low-resolution smart meter technology.