We want to assign a Likelihood of Failure score to each pipe to prioritize our next projects. 

• What are our options? What do advanced analytics -including Artificial Intelligence and Machine Learning- achieve that a utility’s in-house scoring approach does not? (#2 and #3- to be published September 14 and September 21, 2022) 

• Can we trust results from a black box?

• Can we use results from the Physical Condition Assessment of a portion of our system, and extrapolate to the whole system?

• We are a small utility. We have little data. Can we also benefit from advanced analytics?

​

We want to assess Long-term Rehabilitation & Replacement (R&R) needs.

• What’s wrong with using generic Effective Useful Lives to determine long-term R&R needs?

• How do we estimate the Likelihood of Failure of pipes in the future, and leverage those results to optimize our long-term R&R plan?

• Our board set a 1% R&R objective. Will it be enough to control our break rate? Could we do less for a few years? (#1 - to be published September 7, 2022) 

​

We know that the quality of analytical results, such as breaks predictions, depends on the quality of the data.

• Which data is needed?

• What are the typical issues with data and what can be done to improve data quality?

• Is there more to data cleaning than filling data gaps and flagging incoherent values?

• Data quality matters; what about quantity?

• What analytical alternative do we have if we do not have the data (yet), or are a small utility?

​

How does our utility compare with others?

​

How much do the approaches described in the series cost?