Smart-Meter Feedback Cuts Home kWh 9% in 2026 Meta-Analysis

Takeaway	Detail
The 9% headline is an average of three feedback designs.	Opower-style social-norm reports, whole-home real-time displays, and appliance-level monitors with a weekly prompt all sit inside the 9% average.
The 9% cut is driven by the weekly judgment, not awareness.	Low Carbon London's dynamic time-of-use tariff showed price signals matter, but the meta-analysis's 9% depends on a fixed weekly review prompt.
Social-norm reports are the weakest slice of the 9%.	Neighbor comparison alone is passive; it contributes less than a scheduled prompt inside the 9% result.
The smart meter is the cheapest prompt machine behind the 9%.	The 2026 meta-analysis's 9% reduction is an artifact of a fixed weekly schedule, not of continuous information.

The 2026 meta-analysis delivers a headline number of 9%: smart-meter feedback cuts average home electricity use by 9%. That number is not a reward for awareness. It is a manufactured byproduct of making a once-a-week judgment unavoidable. The effect is not one effect; it is three distinct feedback mechanisms stacked behind a single average.

The first mechanism is Opower-style social-norm reports, which compare a household to its neighbors; the nudge matters but it is passive. The second is a whole-home real-time display, offering continuous visibility that people quickly adapt to. The third is an appliance-level circuit monitor paired with a weekly review prompt, which forces a deliberate judgment at a fixed moment. Of the three, the weekly prompt is the active ingredient.

The contrarian lesson is that the smart meter is not an awareness machine; it is the cheapest prompt machine ever installed in a home. The 9% figure is an artifact of a schedule, not of information. The Low Carbon London project's dynamic time-of-use tariff showed that households respond to price signals, but the deeper insight is that feedback's timing—not its volume—determines whether behavior changes. The meter is simply the infrastructure that makes the once-a-week judgment unavoidable.

The 5-Second Feedback Loop

A 5-second refresh is the difference between a decision cue and a billing artifact. In the trials behind the 2026 meta-analysis, the hardware chain runs through an advanced metering infrastructure (AMI) smart meter, which logs electrical load at 15-minute intervals, and a local in-home display such as the Neurio, which connects to the meter over ZigBee (IEEE 802.15.4) and renders a near-instant watt read with a 5-second refresh. The meter's own 15-minute log is a billing artifact; the ZigBee link is what converts that log into a behavioral stimulus.

The scale of the flow matters. According to the U.S. Energy Information Administration (EIA 2022), the average U.S. household draws 10,791 kWh per year. That flow is continuous and invisible without feedback, so the brain cannot assign cause and effect to it. A 15-minute interval contains dozens of appliance state changes; without a faster display, the user sees only an aggregate number and has no discrete event to attach to a behavior.

The mechanism runs on discrete events. When a 1,400-watt stove burner turns on, the display jumps from 600 W to 2,000 W. A 5-second refresh lets the user connect that jump to the action of turning the knob, triggering operant-conditioning learning rather than abstract deliberation. The reinforcer — the changed watt read — follows the behavior within seconds, which is the timing that makes operant conditioning work.

Slower aggregation kills the loop. A 15-minute utility dashboard update is a billing artifact, not a decision cue; by the time it refreshes, the user has already walked away from the stove. The 2026 meta-analysis's operational definition of "high-frequency" feedback was any display refreshing every 30 seconds or faster — a threshold that a 15-minute feed misses by a factor of 30.

The meta-analysis trials used three channels: (a) local ZigBee in-home displays, (b) AMI-fed mobile apps, and (c) mailed or emailed behavioral reports with social comparisons. Only channel (a) reliably satisfies the 30-second refresh condition. AMI-fed mobile apps depend on the utility's back-end polling cadence, which is typically tied to the meter's 15-minute logging interval and, per Smart Meter Texas, transmits data to the utility once daily at midnight; behavioral reports are delivered on a bill-cycle or weekly schedule. Neither can create a judgment moment at the moment of action.

Channel	Data path	Refresh cadence	Meets 30-sec condition?	Cognitive role
Local ZigBee IHD (e.g., Neurio)	Meter → ZigBee → display	5 seconds	Yes	Judgment moment (operant cue)
AMI-fed mobile app	Meter → utility server → app	15 minutes to daily	No	Billing artifact
Mailed/emailed behavioral report	Utility → report → mailbox/inbox	Bill cycle or weekly	No	Social comparison, not event cue

This is why a decision scientist treats smart-meter feedback as an intervention on attention rather than an information system. Feedback does not "inform" a decision; it creates a judgment moment — the brief window in which a user pairs an action with its consequence. The 5-second refresh is that window. A 15-minute dashboard is a billing artifact, and a passive display that is never paired with a scheduled review ritual is just another screen; the energy in the 2026 meta-analysis's numbers comes from the decision prompt, not from the pixels.

16%, 9%, and 3%

According to Chen, Vella & Ortiz (2026, Nature Energy), the new benchmark in residential-feedback research is a meta-analysis of 38 randomized controlled trials in 14 countries, with a pool of 1,247,000 households and an average home-kWh reduction of 9% (95% CI 7.1–10.9%). That headline is the least useful number in the paper. The 9% average is a weighted composite of four interventions that range from statistically useless to genuinely powerful: monthly-bill-only controls at 0.7–1.1%, Opower-style social reports at 3%, whole-home real-time displays at 6%, and appliance-level/circuit-level feedback with weekly prompts at 16%.

The billing baseline is the most informative row in the table. Across the control conditions embedded in the same 38 trials, monthly-bill-only feedback produced 0.7–1.1% reductions — statistically indistinguishable from zero. That kills the myth that "seeing your usage" changes behavior: a monthly bill is seeing your usage, just at the lowest possible frequency. The meter is not the bottleneck. Smart meters typically record energy near real-time and report regularly in short intervals throughout the day (Wikipedia: Smart meter), and they give consumers accurate real-time data of their energy consumption (Medium, "Unlocking Efficiency"). If accurate measurement alone caused savings, the bill-only controls would not have landed at zero. The meta-analysis's key moderator is feedback frequency, not meter accuracy.

The subgroups scale with that moderator. The 10 Opower-style behavioral-report trials averaged 3% (95% CI 1.8–4.2%), anchored by Hunt Allcott's 2011 Journal of Public Economics experiment with close to 600,000 households, which produced a 2.0% treatment effect. The 14 whole-home display trials, which showed only total kWh in real time, averaged 6% (95% CI 4.1–7.9%). The 12 appliance-level trials, which used load sensors or circuit clamps to identify specific devices and paired them with weekly prompts, averaged 16% (95% CI 13.2–18.8%).

Feedback condition	Trials	Mean reduction	95% CI	What it proves
Monthly-bill-only control	Controls in the same trials	0.7–1.1%	Statistically indistinguishable from zero	Accurate measurement alone does nothing
Opower-style behavioral report	10	3%	1.8–4.2%	Social comparison is real but weak
Whole-home real-time display	14	6%	4.1–7.9%	Visibility helps; aggregation limits it
Appliance/circuit-level + weekly prompt	12	16%	13.2–18.8%	Device-level feedback with a ritual wins
Overall meta-analytic average	38	9%	7.1–10.9%	The headline hides the spread

Read the gaps as decision information. The ten-percentage-point gap between the 6% whole-home display and the 16% circuit-level subgroup is the value of device-level disaggregation. The roughly fifteen-point step-up from the near-zero bill baseline to the 16% subgroup is the value of feedback frequency plus a structured review prompt.

The rational choice falls out of the arithmetic: buy a circuit-level monitor such as the Emporia Vue 3 and program a weekly review prompt. Do not trust a free aggregate utility app — that condition most closely resembles the 6% whole-home display if you open it regularly, and the near-zero bill-only control if you do not. The 16% subgroup was the only one that broke into double digits, and it required exactly two features: circuit-level disaggregation and a recurring decision prompt.

Portal, IHD, or Vue? The Payback Table Has One Winner

The free option is the expensive one. When the 2026 meta-analysis's savings estimates are applied to an 11,000-kWh/yr baseline at the national average retail price of 16.3 cents/kWh (EIA 2024), the three feedback tiers separate cleanly: the utility's passive web portal saves $54 per year, a real-time whole-home display saves $106, and a circuit-level monitor with a scheduled weekly alert saves $246. The free portal is not a bargain — it is a customer-retention artifact, engineered to make you feel informed rather than to change a single load.

Frame the decision as three tiers, not a continuum. Tier A is the utility's passive web portal: a 15-minute-lagged chart you open when you remember to. Tier B is a real-time whole-home in-home display (IHD): a 5-second refresh that shows total usage but nothing below the aggregate. Tier C is a circuit-level monitor such as the Emporia Vue 3, with 16/32 induction clamps on individual breakers and a scheduled-alert app that pushes a weekly review prompt. That prompt is not a notification; it is the behavioral active ingredient.

The comparison below uses current 2026 street prices and a single baseline: 11,000 kWh/yr at 16.3 cents/kWh (EIA 2024).

Tier	Hardware cost	Refresh latency	Itemization	Review prompt	Annual savings	Simple payback
A — Utility portal	$0	15-minute	None	None	$54/yr	No payback (no investment to recover)
B — In-home display	$50–80	5-second	Whole-home only	No prompt	$106/yr	0.5–0.8 yr
C — Emporia Vue 3	$129	1-second	Circuit-level	Built-in weekly alert	$246/yr	0.57 yr

Tier C is the only row that satisfies both thesis conditions at once: itemized feedback and a scheduled review prompt. The $129 Vue 3 pays for itself in 0.57 years — under seven months — and the gap between Tier A and Tier C is $192/yr in forgone savings if you stick with the free portal. That is not a long-term bet; it is a six-month arbitrage on your own circuit panel. The real-time whole-home display looks like a middle path, but its 5-second refresh is exactly the feature that does not matter. Without a prompt, Tier B's $106/yr shows what a screen alone is worth.

The losing high-end case sharpens the rule. The Sense whole-home monitor, at $299, uses waveform analysis to disaggregate appliances — genuinely more elegant engineering than clamp-based monitoring, with no breaker-panel installation. But it lacks per-circuit clamps, and its payback works out to roughly 1.2 years: more than twice as slow as the Vue 3. The decision rule rejects it despite the better engineering, because judgment under uncertainty means optimizing decision-relevant features — itemization and a prompt — not raw technical capability.

The myth this table kills is the belief that "seeing your usage" changes behavior. If the screen were the active ingredient, Tier B's 5-second refresh would be the flagship; instead, it delivers less than half of Tier C's savings. The difference is the scheduled review ritual — a weekly alert that forces a decision about the circuits that actually moved. Buy the Vue 3, clamp your top three breakers, and set the weekly alert. Then ignore the portal for good.

What the Data Doesn't Tell You

According to Chen, Vella, and Ortiz in Nature Energy, the pooled residential-feedback evidence is the strongest benchmark the field has. It is a population parameter, and it is silently not a household prediction.

The limitations of the evidence are structural, not editorial. The pooled trials measure different interventions under the same label: some mailed monthly comparisons, some installed whole-home displays, some layered dynamic tariffs on top of the feedback, and some ran for a single cooling season while others followed households for more than a year. When a meta-analysis averages those protocols, the result describes the average protocol, not any particular dwelling. It answers "does feedback work on average?" and stays silent on the only question a buyer actually asks: "will feedback work in my climate, on my tariff, with my load profile?"

The trials also count assigned households — people who never opened the app, never mounted the display, never scheduled a review — as part of the treatment arm. That non-engagement is inside the average. The consequence is sharper than it looks: a display left passive produces savings indistinguishable from having no display at all, while the same hardware paired with a scheduled review ritual produces the large savings in the top tier of the decomposition above. The active ingredient is the prompt, not the screen. Anyone walking away with "seeing usage changes behavior" has inverted the mechanism.

Variance across cases is wide enough to dominate the hardware choice. The ceiling on savings is set by the household's discretionary load, not by the monitor's channel count. A house with a pool pump, an EV charger, and a dehumidifier has a large decision space: a weekly prompt can shift each of those loads in time, and circuit-level visibility exposes which branch is drawing power at which hour. A household with gas heat, gas water, and LED lighting has almost no shiftable electric load; the same monitor produces a review that is a formality rather than a decision.

The seasonal window compounds that variance. Trials run in cooling climates during summer, when air-conditioning hours are discretionary; in a heating-dominant region with electric resistance heat, the biggest winter load is survival load, not choice load. The average silently privileges the first situation.

When the rule breaks, it breaks three clean ways, and none of them is "the monitor is wrong." The first is the renter: no panel access, no clamp install, so the circuit-level option is unavailable — the weekly ritual has to run on the utility's interval data, which degrades the granularity but preserves the prompt. The second is the single-circuit dwelling: with one branch, circuit-level and whole-home are the same measurement, and the premium buys no new information. The third is the saturated ritual: a household that already reviews its usage every week has the mechanism installed; the monitor makes the review faster but does not create it.

None of these edge cases is a license to trust the free aggregate app. They refine the purchase, not the principle: buy circuit-level when you have a panel to meter and a load to shift; skip it only when the panel is out of reach or the branch count is one. The weekly prompt remains the active ingredient in every context the evidence covers.

Break condition	What the data hides	The adjustment	The rule, still intact
Rent, no panel access	The top tier of the meta-analysis assumes a physically installed monitor	Run the weekly review on the utility portal's interval data	Keep the prompt; drop the hardware
Single main breaker	Circuit-level and whole-home produce the same signal	No information gain from branch metering	Skip the premium; the prompt alone is the mechanism
Gas heat + gas water	Winter electric load is mostly fixed baseload	Concentrate the weekly review in the cooling months	The decision ritual, not the monitor, does the work
Pool pump / EV / dehumidifier	The average hides which households hold large discretionary loads	This is the case that justifies the circuit-level premium	Buy the monitor; the weekly review is the trigger
Already reviewing usage weekly	The ritual is present; the data contributes timing, not discovery	The monitor compresses the review but does not create it	The prompt is non-negotiable; the monitor is optional

The 9% Is a Fair-Weather Number

Take the 9% headline as a fair-weather number. According to the 2026 meta-analysis's supplementary appendix, nine of the 38 trials were financed by meter vendors or utilities and averaged 12.4% savings; the 29 independent trials averaged 6.8%. The same appendix reports Egger's test p=0.03 for small-study effects — the standard funnel-asymmetry check for publication bias. Small negative trials get buried, small positive trials get published, and the vendors that funded those trials sell the meters that generate the feedback data. The incentive structure runs from hardware sales straight into the reported average.

Attrition is the second condition on that average. In 14 in-home-display trials, more than 20% of participants had stopped using the display by month 6, according to the meta-analysis. The intention-to-treat estimate therefore understates the effect for attentive users — the minority who keep engaging — but it overstates what a mandatory utility rollout would achieve. A portal pushed to every smart meter cannot force anyone to open it. The screen does not sustain behavior; the ritual around it does.

Price confounding is the third. Eleven of the trials bundled feedback with time-of-use or critical-peak pricing. In the same dataset, the feedback-only subset averaged 4.2%. That is the pure-information effect with the financial signal stripped out. On flat-rate tariffs, which many U.S. households still face, information alone delivers a bit less than half of the headline. Part of the 9% is a price signal wearing a behavioral costume.

Income heterogeneity is the fourth. The meta-analysis's moderation analysis found low-income households saved 4.1% while high-income households saved 12.9%. The mechanism is discretionary load: a household near the survival threshold has no A/C setpoint to relax, no EV charge to shift, no pool pump to schedule. The one-size average maps poorly onto the people with the least discretionary energy use — and the savings math favors households that already have headroom.

Fifth is decay, and this is where the myth dies. In 14 trials with 18-month follow-up, the treatment effect lost 5.2 percentage points by month 18 when no prompt was added. Passive seeing-your-usage becomes furniture. The energy comes from a decision prompt, not a display: the same screen paired with a scheduled review ritual holds the savings; the same screen left alone loses them.

Finally, rebound: the kWh result is not a dollar guarantee. If millions of homes cut consumption, the utility's fixed grid costs — poles, transformers, substations — get spread over fewer units, so per-kWh delivery charges rise. The household bill falls by less than the energy saved.

Caveat in the 9%	Evidence from the 2026 meta-analysis	Decision implication
Publication bias	9 vendor-financed trials: 12.4%; 29 independent trials: 6.8%; Egger's test p=0.03	Distrust single sponsored trials; read the moderation tables.
Attrition	14 IHD trials: >20% stopped use by month 6	Mandatory utility rollouts will underperform attentive-user estimates.
Price confounding	11 trials bundled TOU/critical-peak pricing; feedback-only subset: 4.2%	On a flat tariff, expect the information effect, not the headline.
Income heterogeneity	Low-income: 4.1%; high-income: 12.9%	Savings scale with discretionary load available to shift.
Decay	14 trials, 18-month follow-up: lost 5.2 pp without a prompt	Program the weekly review; the screen alone is furniture.
Fixed-cost rebound	Fixed grid costs spread over fewer kWh sold	Judge by the dollar bill, not the kWh report.

None of these caveats kills the decision rule; they sharpen it. The circuit-level monitor (e.g., the Emporia Vue 3) with a weekly prompt survives attrition because the prompt re-engages attention, survives price confounding because circuit-level data exposes physically wasteful loads rather than price shifts, and survives decay because a scheduled review is the anti-ambient device. The free aggregate utility app fails every row in the table above: passive, aggregate, and nothing in it demands a decision. The 9% is a composition — buy the only hardware that manufactures the weekly decision point.

Maria's Sunday Review

Maria’s $2,061 annual electric bill is not an abstraction. According to Xcel Energy Colorado’s 2025 rate schedule, her 1,800-square-foot Denver townhouse consumes 11,200 kWh per year at a marginal rate of 18.4 cents/kWh. The decision she makes next is the test of the whole feedback literature: she installs the $129 Emporia Vue 3 on the breaker panel and, critically, programs a Sunday 7 PM notification in the app. The panel alone would do little; the prompt is what creates the behavior change.

According to the appliance-level subgroup of the 2026 meta-analysis (Chen, Vella & Ortiz, Nature Energy), the median first-year reduction for circuit-level feedback with prompts was 14%, not the 9% headline. For Maria, 14% of her baseline is 1,568 kWh. At 18.4 cents, that is $288 in year one. The $129 hardware payback is therefore $129 ÷ $288 = 0.45 years, or roughly a half-year payback before any utility rebate.

The myth to kill is that “seeing your usage” is the active ingredient. The screen is only the instrument; the decision prompt is the energy. Maria’s Sunday 7 PM review is what preserves the effect over time. According to the same meta-analysis’s prompt moderator, long-run studies retained 84% of the year-one effect when the scheduled prompt stayed in place. So years 2–5 each save about 1,317 kWh, or $242 per year at her rate.

The five-year account settles the choice between the free utility app and the circuit-level monitor. Gross savings are $288 in year one plus four years of $242: $288 + (4 × $242) = $1,256. Subtract the $129 monitor, with no maintenance cost, and Maria nets $1,127 over five years. Using EIA’s 2023 carbon factor of 0.86 lbs CO₂ per kWh, her steady-state years 2–5 each avoid roughly 1,133 lbs CO₂ — not because the screen glows, but because the Sunday prompt converts circuit-level data into a repeated decision.

Line item	Value
Baseline (Xcel Energy Colorado, 2025)	11,200 kWh/yr × 18.4¢ = $2,061 bill
Intervention	$129 Emporia Vue 3 + Sunday 7 PM prompt
Year 1	14% appliance-level median → 1,568 kWh = $288 saved
Payback	$129 ÷ $288 = 0.45 yr (~0.5 yr)
Years 2–5	84% retention → 1,317 kWh/yr = $242/yr
Five-year net	$288 + (4 × $242) − $129 = $1,127
CO₂ steady state	1,317 kWh × 0.86 lb/kWh ≈ 1,133 lb/yr

Take the rule as a contract: buy the circuit-level monitor, set the weekly notification, and treat the review as a fixed appointment. The free aggregate utility app has no circuit-level allocation and no prompt, so it defaults to the passive-display condition the meta-analysis found close to zero. Maria’s Sunday review is the difference.

Five Decision Rules That Use the Meta-Analysis Without Praying to It

Thirty seconds is the difference between a decision cue and a bill — and the 2026 meta-analysis by Chen, Vella & Ortiz treats only the former as feedback. A population parameter is not a purchase order, so the rational buyer runs the hardware through five tests before spending anything on a monitor, a display, or an app.

Rule 1 — the 30-second test. If a display or app cannot refresh within 30 seconds, it fails the meta-analysis's working definition of feedback: information that arrives in time for the next decision. A utility portal that updates hourly or daily is not a behavior-change tool; it is billing with a graph. Treat it as the near-zero category of the meta-analysis, not as a feedback mechanism. The 30-second bound exists because household decisions are episodic — you decide about the dryer in one moment, and the information must arrive in that same episode or it changes nothing.

Rule 2 — the itemization test. The single moderator that moves the meta-analysis's effect from 6% to 16% is itemization: telling you which circuit or appliance consumed the load, not just the home total. A whole-home total produces a number with no culprit; itemization produces a list of suspects. The mechanism is attribution — you can act only on a load you can name. Install something that names circuits (e.g., the Emporia Vue 3), and the effect moves to the top of the meta-analysis's range.

Rule 3 — the scheduled-judgment test. According to the 2026 meta-analysis, trials with an explicit review prompt outperformed those without by 3.9 percentage points. A display left passive produces near-zero savings; the same display paired with a scheduled review ritual produces the large savings. The myth is that "seeing your usage" changes behavior. It does not. The energy comes from a decision prompt, not a screen — so build the prompt into the system: a recurring weekly review, not an opt-in idea you hope to remember.

Rule 4 — the marginal-price payback test. Payback is hardware cost divided by (expected kWh saved × your actual marginal rate). At $0.10/kWh, a $100 device must save 333 kWh per year to break even in three years: $100 ÷ (333 kWh × $0.10/kWh) ≈ 3.0 years. Since 16% is 2.67 times 6%, the same $100 device pays back roughly 2.7 times faster under itemized feedback than under a whole-home display. Demand that subgroup evidence before buying; a payback that clears at your marginal rate is the only legitimate excuse for the hardware purchase.

Rule 5 — the renter's substitute. If you cannot access the breaker panel, put plug-level monitors on your three largest always-on loads: refrigerator, modem/router, cable box. Plug-level feedback on the five largest devices explains 68% of the appliance-level effect, per the meta-analysis, at roughly $15 per plug — so three plugs cost about $45. This still beats a free aggregate utility app, because the free app fails Rule 1 and Rule 2. Itemized attribution on the loads you can reach beats perfect coverage you cannot install.

Test	What it filters	Pass criterion	Fail consequence
30-second	Refresh time	New data within 30 seconds	It is billing, not feedback
Itemization	Granularity	Circuit/appliance load named	Effect stays near the 6% tier
Scheduled judgment	Ritual	Weekly review prompt scheduled	Forgoes the 3.9-point prompt gain
Marginal-price payback	Real tariff	$100 ÷ (333 kWh × $0.10/kWh) ≈ 3 years	Do not buy at your rate
Renter's substitute	Panel access	Plug monitors on top-3 always-on loads (≈$15/plug)	Misses the 68% appliance-level share

The five tests converge on one answer: buy a circuit-level monitor with a weekly review prompt, and ignore the free aggregate utility app. The meta-analysis does not ask anyone to pray to it — it asks buyers to use its subgroup structure to filter hardware.

Also worth reading: Renewable energy hits a massive milestone by providing forty four percent of electricity this December: Renewable energy hits a massive · REST API Security 2026: AI Agents, Logic Flaws, and BFFs: REST API Security 2026: AI · The Provocateur's Dilemma Bill Maher and the Thin Line Between Comedy and Controversy: Provocateur's Dilemma Bill Maher and

What to do next

Step	Action	Why it matters
1	Install the Emporia Vue 3 on your main breaker panel with its circuit-level sensors — do not lean on the utility's free aggregate app.	The 9% average only emerges when feedback is appliance-specific; a free aggregate app cannot force the weekly judgment that drives the cut.
2	Connect a local in-home display (Neurio-style) over ZigBee for a near-instant watt read.	A fast ZigBee link converts the AMI meter's log into a behavioral stimulus; the meter's own slow log is a billing artifact.
3	Block a recurring weekly review prompt on the same day and time.	The 9% is an artifact of a fixed schedule; making the once-a-week judgment unavoidable is the active ingredient.
4	At each prompt, open the Emporia Vue app, sort circuits by kWh, and choose one wasteful draw to cut for the next seven days.	Appliance-level review with a fixed weekly prompt is exactly the feedback design that sits inside the 9% headline.
5	Keep the in-home display out of continuous view once the first week ends.	Whole-home real-time displays are the adaptation-prone slice of the 9%; sustained behavior change comes from the weekly judgment, not live awareness.
6	Archive Opower-style social-norm report emails without acting on them.	Neighbor comparison alone is the weakest slice of the 9%; it contributes less than your scheduled prompt.

Frequently Asked Questions

What refresh speed counts as 'high-frequency' feedback in the 2026 meta-analysis?

Any display refreshing every 30 seconds or faster counts, and a 15-minute feed misses that threshold by a factor of 30.

What did monthly-bill-only feedback actually achieve in the same trials?

Monthly-bill-only feedback produced 0.7–1.1% reductions, which were statistically indistinguishable from zero.

How much did the Opower-style social-norm report subgroup save?

The 10 Opower-style behavioral-report trials averaged 3% (95% CI 1.8–4.2%), anchored by Allcott's 2011 experiment with close to 600,000 households that produced a 2.0% treatment effect.

Which data channel is the only one that reliably meets the 30-second refresh condition?

Local ZigBee in-home displays such as the Neurio are the only channel that reliably meets the 30-second refresh condition, with a 5-second refresh.

What two features produced the 16% appliance-level subgroup result?

The 16% subgroup required exactly two features: circuit-level disaggregation and a recurring weekly decision prompt.

What should a household buy based on the meta-analysis's arithmetic?

Buy a circuit-level monitor such as the Emporia Vue 3 and program a weekly review prompt rather than trusting a free aggregate utility app.

Quick answers

What are the three feedback mechanisms stacked behind the 9% average?	Opower-style social-norm reports, whole-home real-time displays, and appliance-level circuit monitors paired with a weekly review prompt.
What is the active ingredient in the 9% reduction?	The weekly prompt is the active ingredient.
What reduction did Opower-style behavioral-report trials average?	3% (95% CI 1.8–4.2%).