Buy the second smart meter for the feeder, not for the bill
Ontario's first smart meters were bought to change how customers used electricity and were rarely used to run the grid. The replacement generation should be justified, sequenced and governed as an instrument for the low-voltage network, with the data rules published before the first meter goes on the wall.
Distributors across Canada are now replacing the smart meters they installed around 2010, and most are writing business cases that resemble the first ones: lower reading costs, better bills and customers who will change their habits once they see their data. My position is that a second-generation meter programme should be justified mainly as an investment in seeing the low-voltage grid, sequenced by the needs of individual feeders rather than the efficiency of an installer's routes, and governed by data rules published before the first meter is fitted. I hold that view partly because I managed a first-generation rollout and watched what its data was not used for.
This is not an argument against time-of-use prices, customer apps or the convenience of remote reading. Those benefits are real, and some customers use them well. It is an argument about what should carry the case. A distributor that buys meters for the bill will install them in the order that suits the bill, collect data at the interval that suits the bill and store it where the billing team can reach it. A decade later it will discover, as many of us did, that it owns a sensor on every customer's wall and still learns about most outages from the telephone.
What the first generation was bought to do
Between 2007 and 2009 I ran the smart meter programme at a municipally owned distributor in southern Ontario. We installed 96,000 meters on schedule and under budget. The provincial policy behind them was clear about its purpose: meters would allow time-of-use prices, and time-of-use prices would shift consumption away from peak hours. Our business case followed that purpose. The meters recorded hourly consumption, sent it once a day and reported almost nothing else that our engineers used.
The shift in consumption turned out to be modest. Customers moved some laundry and dishwashing, but the large evening and summer peaks were driven by heating, cooling and cooking that people were not inclined to reschedule for a few cents. Meanwhile the grid itself barely benefited. Our control room could not see voltage at a customer's connection, our planners estimated service-transformer loading from billing totals and nameplate ratings, and our outage maps were drawn from customer calls. We had an instrument at every connection and used it to produce invoices.
I do not say this to criticise the policy makers. In 2006 the communications and data systems needed to use meter data operationally were expensive and immature, and few distributors were asking for them. The point is that the same omission would be inexcusable now. The electrification of heating and transport is changing loads on residential streets faster than any previous change I have seen, and it is happening transformer by transformer.
Sequence the swap by what the feeder needs
At Quillmere we are replacing 287,000 meters over five years, from 2025 to 2029, for C$112 million. Our supplier offered a three-year route-based programme for C$104 million, and our chief financial officer preferred it. I argued for sequencing by feeder: first the meter lots that had failed sampling under the federal seal rules, then the feeders with the fastest growth in heat pumps and vehicle chargers, the most voltage complaints or the poorest outage visibility. Crews move between areas, so the unit cost is higher. Our chief executive agreed on condition that deferred spending would exceed the premium by our 2027 rate application.
Eighteen months in, the evidence is encouraging but incomplete. The new meters report fifteen-minute consumption, voltage and a last-gasp message when supply fails. On the 214 feeders converted so far, voltage data identified 2,140 customers outside the normal range, most of whom we corrected with transformer tap changes rather than new equipment. Measured loading has deferred C$19 million of transformer replacements that our nameplate estimates would have triggered. Outage detection for metered customers now takes about six minutes instead of thirty-eight. The premium we have paid so far is C$3.8 million.
There are limits. Last-gasp messages help on a quiet night and are less useful in an ice storm, when the mesh network itself loses power. Voltage data identifies a problem but not always its cause. And the deferrals depend on our engineers trusting the data enough to leave a transformer in service, which took several months and a few site visits to establish. A distributor without the engineering capacity to act on meter data will not capture these benefits whatever order it installs in.
Publish the data rules before the meter arrives
Fifteen-minute data reveals a great deal about a household: when people wake, when they are away, sometimes what equipment they run. Customers are entitled to know what a distributor will do with it before the meter appears. We made a mistake here. Our statement on the use of interval data was published three months after the first installations, after a councillor asked a question we should have anticipated. The statement now says which data we collect, at what interval, for which grid purposes, who inside the company can see household-level data, how long we keep it and what we will never do with it without consent, including selling it.
The rules matter for a practical reason as well as an ethical one. Engineers want more data at shorter intervals; customers and regulators want less. A distributor that settles those questions in advance can design its systems once. One that settles them in response to complaints rebuilds its data platform every few years.
I would put three questions to any distributor's board considering a replacement programme. First, which planning, operating and restoration decisions will this data change, and how will we know if it does not? Second, does the installation sequence follow the grid's needs or the installer's convenience, and what is the premium for the former? Third, have we told customers, in plain language and before the first visit, what we will and will not do with their data? If a business case cannot answer the first question in specific terms, it is a billing system with a larger budget.