Skip to content
An official website of the OECD. Find out more
Created by the Public Governance Directorate

This website was created by the OECD Observatory of Public Sector Innovation (OPSI), part of the OECD Public Governance Directorate (GOV).

How to validate authenticity

Validation that this is an official OECD website can be found on the Innovative Government page of the corporate OECD website.


Water networks are confronted with aging infrastructure, increased urban population, and climate change. The City of Greater Sudbury has collaborated with CANN Forecast to implement InteliPipes, an AI-based decision-support system that leverages various data sources to improve the overall reliability of its water network. With it the City can (a) understand better the network’s degradation over time, (b) tailor inspection plans and replacement programmes, and (c) optimise watermains replacement.

Innovation Summary

Innovation Overview

In order to develop an efficient life-cycle management strategy for water infrastructure assets and maintain good levels of service, water utility managers are faced with the challenge of making decisions that not only prevent pipe failures whenever possible but also minimize the consequences of these failures. Consequently, an effective risk management decision support system should integrate both likelihood and failure (LoF) and cost of failure (CoF) of linear assets to improve the overall reliability of the water network. However, predicting each of these components accurately is a task that presents several challenges.

On the one hand, the likelihood of failure of a given watermain depends not only on its physical and structural characteristics, but also on environmental and operational factors. Given the very high number of parameters that can reduce the useful life of a watermain, a computational, AI-based approach can leverage existing water infrastructure and break historical data to identify groups of pipes that are most at risk of failure. On the other hand, cost of failure estimation has traditionally been a highly subjective endeavor, which can be attributed to the fact that - direct and indirect - economic, social, and environmental costs are usually difficult to quantify and compare. However, by combining historical work order data and domain knowledge from municipal staff, it is possible to build a data-driven model that provides the most up-to-date insights regarding the potential socio-economic impacts of a future break, for all linear assets.

The City of Greater Sudbury has made investments in the past to collect data on its water infrastructure, watermain breaks and work order history. As such, these datasets can now be used in conjunction with Machine Learning algorithms to identify the most critical pipes in terms of overall risk, combining their likelihood of failure and their cost of failure. The use of innovative, data-driven decision-making tools will help ensure that future investments have the greatest positive impact while limiting adverse consequences related to watermain breaks.

Innovation Description

What Makes Your Project Innovative?

InteliPipes is an innovation because the first step of the process was to run the data through an extensive semi-automated data quality control algorithm designed with best practices provided by experts from Quebec’s Center for Urban Infrastructure. As highlighted by the discipline of Data-centric AI, investing efforts to improve the dataset's quality is crucial in increasing the reliability of future forecasts. The Likelihood of Failure component is based on an unsupervised learning algorithm customized for water infrastructure data after two years of research in partnership with academic partners such as McGill University. The algorithm outputs a decision tree that is easy to understand, making AI explainable, which is a crucial step to help its adoption as a decision-support tool for the public sector. The Consequence of Failure component implemented is an innovative process that combines historical data from the municipality, AI, and expert knowledge.

What is the current status of your innovation?

On the one hand, InteliPipes was able to identify 25Km (2.8 % of the total network) with a break rate of more than 70 breaks/100km/year which is above 10 times the normal rate of failure for the City of Greater Sudbury. On the other hand, the Cost of Failure component of the project was able to leverage expertise and experience from municipal managers and employees across several departments, alongside various data sources to define a data-driven, yet human-centric, consequence of failure evaluation.

Innovation Development

Collaborations & Partnerships

The City of Greater Sudbury Team provided water infrastructure, watermain breaks and work order history. As well as the road class and estimated traffic counts and zoning type from Sudbury Open Data portal. Expert opinion regarding the preferred weighting of each factor in the multi-level Cost of Failure division tree was also a key input from the team. CANN Forecast Team brought AI and data analysis expertise, data collection, quality control and process industrialization. They also provided the likelihood of failure analysis, Cost of Failure analysis and dynamic dashboard.

Users, Stakeholders & Beneficiaries

There are several beneficiaries of this innovation:

  • The City of Greater Sudbury now (a) can have a better understanding of the network’s degradation over time, (b) has tailor inspection plans and replacement programs, and (c) can optimize watermain investments
  • Citizens will ultimately have a more optimized use of their tax dollars and an improved level of service
  • CANN Forecast Team was able to test its innovation and demonstrate that human-centric AI can be successfully implemented in a municipal level

Innovation Reflections

Results, Outcomes & Impacts

The data-driven decision support tool is now being used by several City of Greater Sudbury departments to better understand their water network degradation over time, and optimize inspection plans and replacement programs. The human-centric approach to implementing this AI-based solution had two positive impacts. On the one hand, it helped raise awareness about the importance of quality control and standardization of the datasets. On the other hand, since expert knowledge from municipal staff is directly integrated in the building of the multi-level Cost of Failure decision tree, Sudbury team has brainstormed internally to find more efficient ways to combine economic, social, and environmental costs of infrastructure failure and have shared their insights with CANN Forecast team. This open innovation process is part of a continuous improvement strategy between the City and CANN Forecast in order to serve the population better.

Challenges and Failures

The main challenge encountered was in the process of data collection, quality analysis, and control. Indeed, in almost all municipalities, even the most innovative ones, water infrastructure data is not 100% reliable. This is because these assets have usually been built many decades ago, at a time when IT databases were not readily available. Furthermore, water assets are usually buried underground so it is expensive to double-check missing and uncertain values in the data. This is always a challenge because data errors can flow through the process and ultimately lead to suboptimal or poor decisions. These challenges were addressed by using CANN Forecast’s semi-automated data quality control system designed with best practices provided by Quebec’s Center for Urban Infrastructure experts. The algorithm automatically identified potential human errors, spelling mistakes, misclassifications and inconsistencies in pipes installation date, diameter, material, and break history datasets.

Conditions for Success

The following two conditions are necessary for the success of this innovation:

  • Supporting infrastructure and services: The municipality or public utility should have a GIS layer of its water network and historical records of past repairs for at least the last three to five years
  • Leadership and guidance: The presence of municipal managers capable of pooling expertise and interest across several departments such as Water, Public Works and GIS to ensure that the innovation is integrated in the utility’s business process sustainably

The following conditions are optional, but can greatly help:

  • Policy and rules: The presence of an open data policy provides a very good incentive to produce and share quality data
  • Personal values and motivation: Values of continuous improvement and communication across departments can greatly help reduce silos and accelerate innovation


This innovation has been replicated in several Canadian municipalities. Notably, it has been successfully tested with the Regional Municipality of Peel in Ontario (Canada). By leveraging this innovative approach, the Region of Peel was able to identify 18 km of linear assets - representing 0.39% of the 4,580 km of water mains that constitute the entire water network - that have experienced an average break rate of more than 82 breaks/100 km/yr and a yearly likelihood of failure between 21% and 41% during the 2016-2020 period. As part of an open-innovation framework, CANN Forecast and Peel Region have collaborated to publish a scientific paper regarding the implementation of this innovation for the American Society of Civil Engineers library:

In our opinion, there is great potential for the innovation to be further replicated in the future across the world.

Lessons Learned

There are a couple of lessons from our experience we can share with others like us. It is important to work closely with municipal innovators at every stage of the project, from planning to continuous improvement, as they are the people who understand the business needs. Even if we are technology builders, our ultimate goal is not the tech, but to make sure that what we are building can sustainably integrate itself into the day-to-day processes of our customers and solve their problems. Achieving this may require additional research and development to make AI-based solutions interpretable.


  • Implementation - making the innovation happen
  • Evaluation - understanding whether the innovative initiative has delivered what was needed
  • Diffusing Lessons - using what was learnt to inform other projects and understanding how the innovation can be applied in other ways

Innovation provided by:

Date Published:

29 July 2023

Join our community:

It only takes a few minutes to complete the form and share your project.