Having explored a variety of solutions and met with several potential vendors, Jason selected Rezatec’s Pipeline Risk solution. “With a new technology like AI, it’s important to build trust and confidence that it works and is reliable,” explains Jason. “We needed to prove that Rezatec’s models do work – and that they can genuinely provide us with the accuracy we need and expect to bring precision to our planning.”
Setting the measure for success
“Rezatec needed to predict at least 70% of breaks in the top 30% highest risk mains to hit our benchmark for success,” continues Jason. “The team was positive this could be achieved and were keen to work with us to prove it.”
“We talked to many vendors who are building big models containing as much utility data as they can get to compare against ours, but Rezatec was different because they bring their industry and data science expertise to only our data. That means we get best-in-class analytics, refined through the experience of working with dozens of water utilities across many different scenarios, applied to our own data sets which are enhanced by additional variables through satellite data. It’s a powerful combination.”
Validating geospatial AI
WaterOne began with a validation project to substantiate Rezatec’s risk model on an initial 250 miles, or 10% of the distribution network, which has a good variety of age and materials plus a relatively high break rate. The model was built using 3 years of historical mains break data from 2019 to 2021 and it used 2022 data for validation. The fourth year of data from 2023 was withheld, because it was important to check back to the model as 2023 breaks came in to provide an almost real-time additional layer of validation.