Australia’s critical water pipes break on average 7, 000 times per year. Being able to accurately identify which pipes are at risk of failure will potentially save Australia’s water utilities and the community up to $700 million a year in reactive repairs and maintenance. However, ranking these water pipes according to their calculated risk has mixed results due to their different types of attributes, data incompleteness and data imbalance. This paper describes our experience in improving the performance of classifying and ranking these data via local metric learning. Distance metric learning is a powerful tool that can improve the performance of similarity based classifications. In general, global metric learning techniques do not consider local data distributions, and hence do not perform well on complex/heterogeneous data. Local metric learning methods address this problem but are usually expensive in runtime and memory. This paper proposes a fuzzy-based local metric learning approach that out-performs recently proposed local metric methods, while still being faster than popular global metric learning methods in most cases. Extensive experiments on Australia water pipe datasets demonstrate the effectiveness and performance of our proposed approach.