Publish the restoration time you believe, or publish none

When the power goes out in a storm, the first question from almost every customer is when it will come back. Utilities have become very good at answering quickly. Outage maps update within minutes, text messages go out automatically, and estimated restoration times appear before a single crew has seen the damage. My position is that in major events this speed has become a fault. An estimated restoration time becomes a promise the moment a customer acts on it, and a network should not publish one until it has seen enough of the damage to believe it. In the first day of a major event the honest answer is often that we do not yet know, and we should say so plainly.

How an estimate becomes a promise

In December 2021 a storm brought down lines across part of our network in the Southern Tablelands. Our systems produced estimated restoration times automatically, from the number of customers affected and historical averages for outages of similar size. We published them within hours. We met 41% of them.

The consequences were not abstract. A dairy farmer near Crookwell hired a generator for two days because our estimate said his power would be back that evening, and then had to hire it again for another three. A woman caring for her mother arranged a week's leave around a time we had given her and then extended it twice. A small motel turned away guests on the strength of an estimate that proved too pessimistic. Each of those people made a reasonable decision based on what we had told them. Our call centre spent the week apologising, and our field crews heard about it at every stop.

The problem was not that our estimates were imprecise. Any estimate in a storm is imprecise. It was that we presented guesses as times, produced by a model that could not see broken poles, flooded access tracks or trees across the line in gullies no patrol had yet reached. A customer reading 6 pm Thursday on our website had no way of knowing that the figure came from a formula and not from a crew.

The first day without a number

When a larger storm crossed the Southern Tablelands and the eastern Riverina in November 2024, leaving 96,000 customers without power, I decided that for the first 24 hours we would publish no restoration times at all. Every affected area showed a status of assessing damage, with a commitment about when the first estimates would appear. Our customer and communications team disagreed strongly. They argued that silence would be read as incompetence, that customers would rather have an imperfect time than none, and that the shareholding ministers' offices would insist on a number. They were right about the ministers' offices. The first day was difficult.

What we gave customers instead was a schedule. We told them that patrols, including two helicopters, were inspecting every affected feeder, that the first estimates would appear by 7 pm the next day, and that we would update them at 7 am and 7 pm each day after that. We published our restoration priorities as a list: hazards to life; then hospitals, water and sewage pumping and telecommunications; then the feeders where each crew-hour would return the most customers; then isolated lines. People could see where they stood before they had a time.

When estimates appeared, they were ranges by town rather than times by street: Thursday or Friday, not 6 pm Thursday. We told the farms at the ends of single-wire lines plainly that they would be among the last and that it might be more than a week. Some were angry. Several told us later that the plain answer let them make arrangements they would not otherwise have made.

We met 74% of the ranges we published. Ninety per cent of customers had power within four days, and the last were reconnected on day eleven.

I should answer one criticism directly. A range can look like a hedge, a way of never being wrong. Ours were not open-ended: no town's range was wider than two days, and a range counted as missed if a single street in that town was still off at its end. Measured that strictly, 74% is not a flattering number. It is still a long way from 41%, and the difference is that the people who set the ranges had seen the damage.

What the control room changed

The change was not mainly in communications. It was in how the control room and the field produce an estimate. We now require a damage-patrol report on a feeder before its estimate is published, and the estimate is signed by the regional field manager who will have to meet it, not issued by the outage management system. The system still proposes a figure, but it is treated as a draft. The field manager can widen it, and often does.

We also publish our hit rate after every major event, by region, in the same report that goes to the board's safety and network risk committee. That has changed behaviour more than any instruction. A field manager whose estimates are published and measured is careful about what they sign.

There are fair objections. In small, routine outages an automated estimate is usually close enough, and we still publish one immediately; our rule applies only when an event exceeds a defined size. Some customers genuinely prefer a precise time, even one that may be wrong, and our approach disappoints them. And holding estimates for a day puts heavy pressure on the call centre, which needed more people and better scripts than we gave it in 2024.

I think the principle reaches beyond storms. Any operator that tells the public when a service will return, whether it runs a water network, a railway or a telecommunications system, is making a promise that people will act on. The responsible course is to publish only what the people accountable for meeting it are prepared to sign, to say clearly when nothing can yet be said, and to report afterwards how often the promise was kept.