How Smart Meters Could Transform Electricity Bills

New research shows that smart electricity meters could double utility profits while simultaneously reducing energy bills for low-income households, if utilities use them right.

A dynamic pricing model for designed by Bora Keskin of Duke University’s Fuqua School of Business, Yuexing Li of Johns Hopkins University, and Nur Sunar of the University of North Carolina at Chapel Hill, shows how pairing smart meters with dynamic pricing turns real-time data into a win-win for companies and consumers.

In a paper published in the journal Operations Research, Keskin, Li, and Sunar developed a model that segments customers using fresh data streams, allowing utilities to tailor dynamic rates to each group.

The model identifies groups of households that respond similarly to weather conditions, home features, and shifting electricity prices, enabling utilities to customize dynamic rates for each segment.

“We wanted to understand how utilities could move beyond flat rates,” Keskin said. “The data from smart meters and from ad hoc customer surveys showed us a path to smarter pricing.”

Smart meters and dynamic pricing

Smart meters are quickly becoming a staple of the modern grid. By 2020, 75% of U.S. households already had one smart meter, and similar adoption levels are projected across Europe. 

These devices collect real-time consumption data, and some advanced meters even allow utilities to communicate prices to consumers. 

Smart meters are essential to dynamic pricing, Keskin said. At the very basic level, dynamic pricing adjusts electricity prices by hour of day or season, with higher rates in the summer and during business hours.

U.S. utilities have rolled out dynamic pricing programs, and some—like Illinois-based Commonwealth Edison—now offer customers real-time hourly rates. From 2013 to 2019, the number of Americans on time-varying plans grew by more than 90%. European regulators have gone further: EU law requires at least one energy supplier in each member state to offer real-time pricing.

Designing smarter electricity rates

Despite growing enthusiasm, creating truly effective pricing plans is hard, Keskin said. Utilities must understand how sensitive each household is to price changes—a mostly unknown factor, the researchers point out. Traditional data, often gathered under flat-rate pricing, offers little variation from which to learn.

Keskin and co-authors designed a model that learns on the fly how the many variables affecting electricity demand can be used to identify an optimal price. Because smart meters provide highly granular information, they enable customer segmentation and personalized rates using a wide range of inputs—such as home size, appliance usage, and weather conditions—as well as unpredictable factors like equipment issues and maintenance events.

To make sense of these diverse inputs, the model organizes them into categories based on whether they vary over time (such as weather patterns), across customers (like home size or number of appliances), or across both time and customers (such as maintenance issues or HVAC failures).

This structure allows the algorithm to isolate the influence of each variable on demand. 
The researchers then paired this structure with a clustering technique—called “spectral clustering”—that segments customers into groups that behave similarly. Once those clusters are identified, they designed a separate model for each group to determine the profit-maximizing price. 

“When customers differ in so many ways at once, you need a tool that sees the structure beneath the noise,” Keskin says. “Clustering turned out to be ideal.”

For practical reasons, the researchers calibrated their model for utility profit maximization, but the same design can support different objectives, such as reducing electricity loads or improving consumer welfare, Keskin said.

A real-world test: The Austin energy case study

With the model developed, the researchers tested it using real-world data from Austin Energy’s pricing trials.

The utility had experimented with price incentives, allowing the researchers to learn the price sensitivity within each cluster.

The researchers found that their model would increase Austin Energy’s profits by more than 100% over three months, compared with the utility’s historical pricing decisions.

To examine how dynamic pricing affects low-income households, the researchers also tested their model with a narrower range of four price points, far less varied than the utility’s historical set of prices. They found that this strategy would charge low-income households a lower average price than higher-income customers, all while improving utility profits.

“By personalizing the prices for different customers, utilities can generate more demand when the customers need it, and customers benefit from the lower prices,” Keskin said. “It turns out to be a win-win for both.”

What utilities can learn from this research

Keskin’s research offers utilities a practical blueprint for adopting customized dynamic pricing.

The first step is price experimentation, he said, which involves deliberate price trials that reveal how customers adjust their consumption. In Austin Energy’s case, those trials were what made it possible to measure price sensitivity in the first place, and any utility hoping to personalize prices will need some version of that early learning phase, Keskin said.

But experimentation has limits, he said. Utilities must “balance learning and earning,” gathering enough information to understand demand before shifting from exploration to profit generation—the researchers even provide an algorithm that optimizes this initial process for utilities.

Once the learning phase concludes, the model uses clustering to identify distinct customer segments, then builds a tailored demand model for each segment. This method lets utilities set prices that reflect meaningful differences in customer behavior rather than treating all households as identical.

“If utilities experiment wisely and segment smartly, the data will do the heavy lifting—turning pricing from trial and error into a strategy,” Keskin said.