Do Smart Thermostats Really Save Money? The Honest Numbers

Yes, but the honest number is roughly 8 percent of a heating and cooling bill, not the 20-something percent that appears in product advertising. That 8 percent is the figure published by the federal program that actually certifies these devices, and it is measured against the mix of thermostats already installed in American homes. For a lot of households it works out to less than the price of the thermostat over the first two years.

The number ENERGY STAR actually publishes

EPA introduced the smart thermostat label in 2017. In its own FAQ document, the program states that for certified smart thermostats, savings are approximately 8 percent of heating and cooling bills, or $50 per year. The same two figures appear in the program technical bulletin on certified smart thermostats, a document whose supporting data is dated to the end of 2017.

Two things about that number matter more than the number itself.

First, the comparison. ENERGY STAR does not measure a smart thermostat against a plain dial on a wall. Savings are measured against the installed base of all other thermostats in the United States, which already includes millions of seven-day programmable units that people set up sensibly and left alone. If that describes your current thermostat, you have already captured part of the benefit, and your own saving will be smaller than the published average.

Second, the arithmetic behind $50. Eight percent of what? Divide $50 by 0.08 and you get $625, which implies a heating and cooling bill of about $625 a year in the underlying data. The ENERGY STAR consumer page for smart thermostats says the average American household now spends more than $900 a year on heating and cooling. Apply the same 8 percent to $900 and the saving is $72, not $50. The percentage is the durable part of the claim. The dollar figure is a snapshot of prices from nearly a decade ago, and it should be read that way.

What the certification threshold is, and what it is not

The key product criteria are more revealing than the marketing summary. To carry the label, a model must demonstrate an annual heating runtime reduction of at least 8 percent and an annual cooling runtime reduction of at least 10 percent, each measured at the lower 95 percent confidence limit. It must also deliver at least 4 percent heating and 5 percent cooling savings at the 20th percentile of the homes in the test sample.

That last clause is the one worth reading twice. The 20th percentile requirement exists because the program knows that roughly one home in five will do markedly worse than the average. A certified thermostat is permitted to deliver as little as 4 percent heating savings to the bottom fifth of the homes using it and still keep the label. The certification is a floor for a population, not a promise to a household.

Verification is also unusual for ENERGY STAR. Rather than a laboratory test, the program collects a year of operating data from a sample of homes across the country, aggregates it with EPA-supplied software, and sends the result to a third-party certification body. That is a genuinely stronger method than an engineering estimate. It is still an average across a sample that is not your house.

One further caveat on currency. These criteria come from Version 1.0 of the specification. EPA has been working on a replacement since 2021, and as of September 2026 the Version 2.0 specification is still at Draft 2, released in November 2024. Every smart thermostat certified today is tested against rules written in 2017.

Where the bigger numbers come from

Savings claims in the high teens and twenties generally come from a manufacturer analysing data from its own customers. Those studies are not fabricated, but the sample selects itself. People who buy and install a connected thermostat are people who already care about their energy bills, and their prior behaviour is not the national baseline.

The Department of Energy has been testing these claims against building simulation instead. A 2023 Building Technologies Office peer review presentation from the National Renewable Energy Laboratory on simulation-driven smart thermostat benchmarking sets the contrast out directly: manufacturer literature claims of 24 to 28 percent, against simulated results in which most homes saved 0 to 3 percent from a basic algorithm such as occupancy sensing alone, and 5 to 8 percent once occupancy sensing and nighttime setback were combined. Those are preliminary modelling results that were still being validated against field data when presented. They are not the last word. But they point the same direction as the 8 percent, and the opposite direction from the advertising.

Work out your own payback

You need two inputs: your annual spend on heating and cooling, and the all-in cost of the device after rebates.

Most people guess the first one wrong. EIA reports that space heating and air conditioning together accounted for 52 percent of household energy consumption in 2020, but the split varies enormously by region and by system type. Pull twelve months of bills, subtract a summer shoulder month as your baseline non-HVAC load, and multiply the difference out. A rough number from your own bills beats a precise number from somebody else bills.

Then apply the 8 percent.

The two cost columns are scenario inputs, not published averages. The $150 column stands for a mid-range certified unit that you install yourself. The $350 column stands for a premium unit plus an electrician to run a common wire, which older homes frequently need. Substitute your own quote, and subtract any utility rebate before dividing.

The decision rule

1. Take your annual heating and cooling spend from last year bills.
2. Multiply by 0.08. That is the expected annual saving, using the ENERGY STAR certified average.
3. Multiply that by 5, a conservative service life for a connected device that depends on a manufacturer cloud service.
4. If the five-year total is smaller than your all-in cost after rebates, do not buy it for the energy saving.

Rearranged into a threshold: at $150 all-in you need to be spending about $375 a year on heating and cooling for the device to repay itself within five years. At $350 all-in you need about $875. Below those lines, the purchase may still be worth making for remote control, for equipment fault alerts, or for a utility demand response payment. It is not worth making for the energy.

Four situations where the saving is close to nothing

You heat with a heat pump

This is the big one, and it is the case the marketing almost never addresses. DOE guidance on operating and maintaining your heat pump is explicit: do not set back the heat pump thermostat if it causes the backup heating to come on, because backup electric resistance strips are usually more expensive to operate. A deep overnight setback on a heat pump can cost more than it saves when the system calls for strip heat to recover in the morning.

Somebody is home all day

Setback savings come from hours when the house is empty or everyone is asleep. Occupancy sensing has nothing to work with in a house that is occupied from 7am to 11pm. Remote workers, retirees and households with young children should expect to land in the bottom part of the distribution, not the middle.

Your old thermostat was already programmed properly

DOE advice on programming a thermostat for automatic savings is that you can save around 10 percent a year on heating and cooling by turning the thermostat back 10 to 15 degrees Fahrenheit for eight hours. Its seasonal guidance puts the same manoeuvre at 5 to 15 percent off a heating bill. None of that requires a connected device. It requires a schedule. A basic programmable thermostat costing a fraction of a smart one delivers most of it, which is exactly why the smart thermostat certified advantage over the installed base is only 8 percent and not 30.

You override it

Every saving in every study above assumes the schedule survives contact with the household. A thermostat that is held at 72 degrees all winter because one person in the house is cold saves nothing regardless of what it cost.

Rising electricity prices cut both ways

A fixed percentage saving on a rising bill is worth more in dollars each year. The EIA September 2026 Short-Term Energy Outlook puts the average US residential electricity price at 17.30 cents per kilowatthour for 2025, with a forecast of 18.20 cents for 2026 and 18.59 cents for 2027.

The same rise works against you in a subtler way. As the marginal cost of comfort goes up, the temptation to hold a tighter setpoint becomes more expensive, and the households most likely to feel the price increase are also the households least able to fund a $350 upgrade in the first place.

The program behind the label is being handed over

Worth knowing if you are relying on the certification. Under the ENERGY STAR transition plan finalised on 28 August 2026, primary leadership of the program moves from EPA to the Department of Energy, with a twelve-month deadline for the full transition. The plan states that certification and verification processes remain substantively unchanged. Nothing in it changes the 8 percent criterion. It does mean the Version 2.0 specification, already four years in draft, is now being handed between agencies.

The honest version

A certified smart thermostat is a real product that delivers a real, modest, verified saving. It is roughly 8 percent off the heating and cooling portion of your energy bill, against a baseline that already includes ordinary programmable thermostats, and one home in five will do considerably worse than that. If you spend a lot on heating and cooling, leave the house on a predictable schedule, and do not have a heat pump with electric backup, the payback is short and the purchase is easy to justify. If you already run a sensible schedule on a working programmable thermostat, the energy case is weak and you are mostly buying an app.

The thing nobody selling these devices will tell you is that the largest single determinant of your saving is not the thermostat. It is the insulation, the ductwork and the age of the equipment behind it. A thermostat can only decide when the system runs. It cannot make a leaky house cheap to heat.

General information, current as of September 2026, and not personalised advice.