Chief Technology Officer / United States
Three recommendations that cost something colleagues expected
- Water customer connections
- 126,000
- Plus 41,000 wastewater connections, December 2025
- Water and wastewater systems
- 78
- Southern and central Indiana and northern Kentucky
- 2026 technology budget
- US$29.5M
- US$17.8M capital and US$11.7M operating
Decisions & consequences
01Holding an acquired plant on local control for eleven months
In September 2020 Halverick completed the purchase of the water and wastewater systems of a small northern Kentucky city: a 3.2-million-gallon-a-day treatment plant, a wastewater plant and 6,800 connections. The acquisition filing assumed the plants would be monitored from our central control room within three months, allowing two overnight operator positions to be removed. On our first site visit we found an internet-facing remote-desktop connection to the plant computer, an operating system out of support since 2014 and controller logic that differed from the only backup copy in three places.
We closed the remote connection on the first day. The harder question came next. Our chief operating officer wanted a virtual private network link within ninety days, and finance noted that the staffing assumption had been part of the case presented to directors. I recommended keeping the plant on local control, staffed overnight, until we had rebuilt its network and verified every controller program. That meant holding two operator positions longer than planned and accepting a slower integration than the filing described.
The chief executive agreed, and asked me to put the cost in front of the board myself. Over eleven months we spent US$410,000 on overnight staffing and contract support, beyond the US$1.9 million network and controller rebuild already in the plan. The verification found that a high-level alarm on an elevated storage tank had been disabled in software. Operators restored it and changed their rounds until the rebuild was complete.
The plant joined the central system in August 2021 through a segmented network, using our standard logic library. One overnight post was retired; the other became a roving role covering three acquired sites. The integration report recorded the delay against the original filing.
I would now insist on access to an acquired plant's control system during due diligence, before the price and integration timetable are agreed. The eleven months were justified, but they should have been in the acquisition case from the start.
Verifying district meters before reporting a leakage reduction
In 2022 and 2023 we installed 64 district flow meters across four older systems and subscribed to a leakage-analytics service for US$380,000 a year. After fourteen months the vendor's dashboard showed real losses falling from 61 to 44 gallons per connection per day. Our vice president of operations wanted to present the result to directors in February 2024 and request US$14 million to extend the programme to twelve more systems.
The repairs were real: crews had found and fixed 142 leaks. My concern was the measurement. Several meters had been installed in cramped vaults close to bends and valves, and nobody had tested them against an independent reference after commissioning. I asked for every district meter to be verified with portable insertion meters and for district totals to be reconciled against plant production meters before any figure went to the board. Operations disagreed, reasonably: the verification would cost money, delay the capital request by a planning cycle and imply that their teams had overstated the result.
The verification cost US$640,000, including excavation for access at nine vaults. Thirteen meters were reading low by between 4 and 11 per cent, two of them because the wrong pipe diameter had been entered at configuration. The verified figure was 55 gallons per connection per day: a genuine improvement of six, not seventeen.
The board approved a smaller US$3.1 million extension to two systems with the strongest verified results. We renegotiated the analytics contract so that no new district enters the service until its meters pass a commissioning test. The vendor disputed our findings for several months before accepting the corrected baseline. My relationship with the operations team needed repair as well; I had to make clear that the leak crews' work was never in question.
I would build the commissioning test into the original installation contract, so that no vendor dashboard is the first place a performance figure appears.
Choosing a subscription that removed capital from the regulated plan
Our work and asset management system dated from 2009, its database would lose support in 2026, and crews in several acquired areas still worked from paper orders. In late 2024 we compared two replacements. An on-premises licence with capitalised implementation would cost about US$9.8 million, plus roughly US$900,000 a year in hosting and support. A subscription service would cost US$4.1 million to implement and US$1.45 million a year. Over ten years the totals were close.
The difference lay in regulatory treatment. Capital investment would enter rate base and earn an allowed return; subscription fees would be recovered as operating expense without one, and the treatment of implementation costs varies between commissions. Our chief financial officer and regulatory affairs preferred the licence. Two investor-nominated directors asked what the subscription would do to earnings. In the early years, the forgone return was about US$700,000 annually.
I recommended the subscription. My infrastructure team was patching 140 on-premises servers, the old system had no workable mobile access for field crews, and the vendor would carry upgrades we had repeatedly deferred. I presented both rate treatments side by side, including the risk that implementation costs might not be deferred in both states. The board approved the subscription in March 2025, with regulatory affairs asked to seek permission to defer implementation costs for review in the next rate case in each state.
One commission approved deferral in December 2025. The request in the second state remained open in October 2026. Three of our four operating regions went live in January 2026, retiring paper orders there; the fourth is due in early 2027. Thirty-one servers were decommissioned. Earnings carried the expected reduction, and I did not describe it as a saving.
I would involve regulatory affairs in vendor selection from the first week. The side-by-side comparison should have opened the first paper, rather than arriving in response to a director's question.