A homeowner in Vermont set up a notification on his energy monitor: alert me if the hot water heater hasn't turned on within eight hours. Two weeks later, the alert fired. He checked the monitoring app on his phone. The controller on his hydronic in-floor heating had failed overnight. Because the monitor caught it within hours, the water in the tank was still above 140°F. He called a plumber the next morning, replaced the controller, and the system never dropped below operating temperature. Had the failure gone unnoticed for 48 hours in a Vermont winter, the water temperature in those heating lines could have dropped below freezing, cracking the tubing embedded in his concrete slab. Repair estimate for a failed hydronic floor: $8,000 to $15,000, plus tearing out finished flooring to access the damage.
That homeowner's monitor cost $299.
What a panel monitor actually does
The device in question is a Sense energy monitor, though it is not the only one on the market. Two current transformer clamps wrap around the main service wires inside your electrical panel. No splicing. No rewiring. A licensed electrician installs the hardware in about 30 minutes, connects it to a dedicated 240V breaker, and the monitor begins sampling your home's electrical current at one megahertz. That is one million readings per second, yielding roughly four million data points every second when you account for both voltage and current on each leg of a split-phase residential system.
Your utility's smart meter, by contrast, records one reading every 15 minutes. Sense samples 84 million times more frequently than the device your utility uses to bill you. That sampling density is what makes real-time disaggregation possible. Every electrical device in your home has a unique power signature. A compressor starting up draws current in a pattern that looks nothing like a resistive heating element warming up, and neither of those looks like a brushless motor ramping to speed. A microwave oven has a startup profile so specific that the algorithm can distinguish it from a toaster operating at similar wattage. Your dryer's heating element cycles in a rhythm determined by the thermostat setpoint, the drum load, and the lint accumulation on the filter. These are not abstract patterns but fingerprints, the kind of electrical identity that a machine learning model trained on data from thousands of homes can match to a specific appliance type, sometimes a specific manufacturer.
Twelve devices in the first month. Thirty within a year. MIT Technology Review reported the company's claim of disaggregating 80% of household energy use from a single measurement point, a figure that put it "at the cutting edge" of consumer-accessible NILM (non-intrusive load monitoring) technology.
Fault detection is the real value
Energy awareness is useful. Knowing which outlet draws phantom load is satisfying in a nerdy, watch-the-numbers way that appeals to the kind of homeowner who reads this site. But fault detection is worth money, real money, the kind that shows up as an avoided insurance claim or a compressor that did not burn out because someone caught the problem in time.
Sense's "Motor Stalls" feature identifies when appliances with large motors are repeatedly failing to start. A compressor trying and failing to engage every few minutes is drawing startup current each time without ever reaching steady-state operation, wasting electricity and damaging the motor. One user discovered a failing start capacitor on his heat pump through the Sense app before he noticed any comfort change inside the house. He replaced a $12 capacitor instead of paying $800 for an emergency HVAC call after the compressor burned out.
Water heaters are the most consequential example. A resistive water heater approaching failure often doubles or triples its energy draw for months before it finally ruptures. The element degrades gradually: longer runs to setpoint, more frequent cycles, higher current each time. A healthy 4,500-watt element running 3 hours per day costs about $1.35 daily at the national average residential rate of $0.167/kWh. A degrading element pulling 6,000 watts and running 4.5 hours costs $4.50 per day. That is $1,150 in excess annual electricity cost, quietly accumulating on your utility bill in increments too small to notice month over month but large enough to total more than the cost of a replacement water heater over two years.
A monitor that catches a water heater's degradation six months before catastrophic failure saves the replacement cost ($1,200 to $2,000 for a standard tank unit including installation), avoids the damage claim entirely, and preserves the insurance premium. Total cost: $449, with $299 for the monitor and $150 for installation. Against a potential $4,000 to $12,000 loss, the math is not close. It is not even in the same neighborhood.
What your utility already knows
Panel-mounted monitors like Sense are consumer products. On the utility side, a company called Bidgely has been disaggregating smart meter data at the 15-minute resolution your utility already collects, using an AI platform backed by 16 patents acquired partly through its purchase of Grid4C. Bidgely processes over a terabyte of energy consumption data daily from utility customers worldwide and claims its platform has collectively saved 1.5 terawatt-hours of energy across gas, electric, dual-fuel, and water customers.
Bidgely's real-world deployment results, published in a May 2026 announcement, include a 70% peak load reduction at one utility, 3x improvement in EV program recruitment, and 98% retention on time-of-use rate plans. A major investor-owned utility in the southwestern United States is targeting a 5 to 10x return on investment using Bidgely's UtilityAI Pro platform to extract appliance-level intelligence from its existing smart meter infrastructure.
What this means for homeowners: your utility may already have appliance-level insight into your home's energy behavior, derived from the meter data they collect every 15 minutes, analyzed by machine learning models trained on millions of households. You have probably never seen this data. It exists, and it describes your home. Whether your utility shares it with you depends on whether they have licensed the disaggregation platform, built a customer-facing interface that presents the results in a form homeowners can actually act on, and decided that transparency about what those meters reveal is worth the development cost, which most utilities have not. Bidgely sits atop Guidehouse Research's Home Energy Management leaderboard precisely because utilities are buying the analytics, not because homeowners know it exists.
New construction makes the math even simpler
Installing a panel monitor during new construction eliminates the retrofit premium. The electrician is already on-site, the panel is open and accessible, and adding a dedicated 240V breaker along with the monitoring hardware takes minutes rather than the 30-minute minimum of a separate retrofit visit. Device cost remains $299, but the labor marginal cost drops to near zero because it folds into existing electrical work.
Compare them to what builders already spec. A Nest thermostat: $280 for temperature. An Ecobee Premium: $250, plus a room sensor. Both are standard line items for the move-up market.
A $299 energy monitor tells you which device is costing you money, which one is about to fail, and how your actual energy consumption compares to what the HERS rater predicted. That last point matters for builders pursuing ENERGY STAR or green certifications: the persistent gap between modeled performance and real-world performance has been documented repeatedly, and the primary reason is that nobody measures actual consumption at the appliance level after the certificate is issued. That closes the loop.
Researchers at Purdue University tested a machine learning predictive control system in their DC Nanogrid House, a fully instrumented test home. Over a 40-day winter trial, the algorithm, which incorporated predicted solar radiation, humidity, forecasted temperature, and room occupancy data, reduced energy consumption by 19% compared to a comparable stretch from a previous winter. Dollar savings: approximately $300 for a single household over six weeks. Extrapolated nationally, the team estimated $22 billion per year in savings across all U.S. households.
That extrapolation is aggressive given the evidence: one house, one climate zone, 40 days. But the point is directional: the data generated by panel-level monitoring enables control strategies that thermostats alone cannot execute because thermostats do not know what the rest of the house is doing.
Where this falls short
Sense's own documentation acknowledges that many devices will never be identified. Low-wattage loads like laptops, smart speakers, phone chargers, and game consoles draw too little power to produce a distinguishable signature at 1 MHz sampling. Variable-speed devices, increasingly common in high-efficiency HVAC systems and modern appliances, are harder to fingerprint because their electrical signature changes continuously rather than presenting the discrete on-off transitions the algorithm was designed to detect.
The 80% disaggregation claim comes from Sense itself. Independent academic research on non-intrusive load monitoring, using publicly available datasets like REDD and REFIT, reports accuracy ranges of 81% to 94.6% for home-level disaggregation depending on the algorithm and appliance type, with performance dropping for multi-appliance scenarios and open-set environments where unknown devices are present. A 2026 study published in International Transactions on Electrical Energy Systems achieved 71% to 78% similarity in live disaggregation across two real homes, a meaningful gap from the 80% commercial claim.
Bidgely's utility-side disaggregation operates at 15-minute smart meter resolution, which is orders of magnitude coarser than panel-mounted monitoring. It works because Bidgely trains on data from millions of homes, using aggregate patterns to infer appliance behavior rather than detecting individual power signatures. Accuracy trade-offs at this resolution are real: the system can identify that you have an EV charger and estimate its monthly consumption, but it cannot tell you that your dryer vent is 40% blocked by lint and your drying cycles have lengthened by 12 minutes each.
The Purdue study was a single instrumented house, not a randomized controlled trial across building types and climates. No published study has compared fault detection outcomes in homes with panel monitoring against homes without it in a controlled experimental design, which means the insurance cost avoidance and maintenance savings described in this article, while directionally supported by the available evidence, remain extrapolations from individual cases rather than measured population-level outcomes. The insurance cost data cited in this article reflects industry averages, not actuarial breakdowns by monitoring status. No insurer currently offers a premium discount for homes with energy disaggregation monitoring installed, though several offer discounts for smart water shutoff valves, which address the same underlying risk from a plumbing rather than electrical vantage point.
What to do with this
If you are building a new home, add a panel energy monitor as a line item at $299 for the device and negligible marginal install cost during construction. You will learn more about your home's actual energy performance in the first 90 days than the HERS rating told you, and if a major appliance begins degrading, you will know before it fails catastrophically.
If you own an existing home with a standard residential panel and at least one available breaker slot, a retrofit installation costs approximately $450, covering the $299 monitoring device and roughly $150 for a licensed electrician to mount the current transformer clamps on your main service wires and connect the dedicated breaker. Prioritize the install if your water heater is older than eight years, your HVAC system is older than 10 years, or your electricity bill has crept upward without an obvious explanation. The monitor won't name the problem. It shows the pattern that reveals one.
If you are a production builder, consider this: every smart thermostat you install at $250 per unit tells the homeowner their house is 72 degrees. A $299 monitor in the same home tells them their HVAC system ran 23% more last month than the month before, their water heater is drawing 30% more current than baseline, and their dryer takes 14 minutes longer per cycle than it did six months ago. One of those devices monitors comfort. The other monitors the house.
Limitations
All disaggregation accuracy claims cited here come from manufacturer disclosures or company-funded studies. No independent, peer-reviewed study has validated Sense's 80% disaggregation rate under controlled residential conditions. Bidgely's reported energy savings (1.5 TWh) are aggregate portfolio figures that cannot be attributed to disaggregation alone versus broader utility program effects. The Purdue energy savings percentage (19%) applies to a single test home in a specific climate zone during one winter; reproducibility across housing types and climates is undemonstrated. Water heater failure cost estimates ($4,000 to $12,000) come from insurance industry averages that include high-severity incidents (finished basements, adjacent room damage) and may overstate typical single-story slab-on-grade scenarios where a failed tank drains to a garage floor. The ROI calculation assumes the homeowner acts on the monitoring data; a monitor that sends alerts to an app nobody opens provides no benefit. Installation requires a licensed electrician and a panel with available breaker space, which older homes with fully loaded 100-amp panels may not have without a service upgrade.
← Back to AI Home Building