Measurement & verification
Measurement and verification: how VPP performance gets proven and paid
Every VPP payment rests on an estimate of something that didn’t happen: the power a home would have used without the event. Here is how grid operators and utilities measure it.
VPP Economics has no affiliation with, sponsorship from, or endorsement by any utility, aggregator, platform, or device maker named here.
The short answer
- Every VPP payment rests on a baseline — an estimate of what a home would have used if the event never happened. The reduction is the baseline minus what the meter or device actually recorded (U.S. Department of Energy).
- Grid operators build baselines differently. PJM uses the highest 4 of the last 5 weekdays; NYISO the highest 5 of 10; ISO New England and CAISO average 10 days, with different adjustments (PJM, NYISO study, ISO-NE, CAISO).
- Homes are measured differently from factories. Residential programs often use control groups, measured average values — Texas uses 1.40 kW per smart thermostat — or data from the device itself (Texas PUC).
- So the same event can pay differently depending on the rule. Calculator estimates, including ours, are a starting point; settlement is what counts.
How a baseline turns into a payment
Illustrative day for one home with a smart thermostat and a 5–8 p.m. event.
Swipe to see the full chart →
Six ways performance is measured
North American Energy Standards Board (NAESB) standards, adopted by FERC for wholesale demand response, name the main methods; the U.S. Department of Energy’s guide adds randomized control groups (DOE, NAESB).
Baseline from interval meter
Uses the site’s own past interval-meter data (sometimes with weather) to estimate what it would have used. The standard for metered homes and businesses; NAESB calls it Baseline Type I.
Meter before / meter after
Compares usage just before an event with usage during it. Works for fast, short dispatches; ERCOT’s aggregated-DER pilot uses it for some resources.
Statistical sampling
Estimates a whole fleet from a metered sample when not every home has interval data (Baseline Type II). ERCOT’s pilot does not allow it.
Metering device output
Pays for what the device itself delivers, measured at the device. Originally written for backup generators; it fits batteries well.
Maximum base load
Pays a site for staying below a pre-agreed level during an event, regardless of what it would have used.
Control group
Leaves a random set of enrolled homes out of each event and compares the two groups. CAISO requires at least 150 customers in the control group.
How each grid operator picks baseline days
Which recent days set the baseline
Each square is a recent non-event weekday the rule looks at; filled squares are the days it uses.
Swipe to see the full chart →
Choosing the highest days, as PJM, NYISO, and CAISO’s residential rule do, gives a higher baseline and more credit for the same behavior than a straight average. Texas’s ERCOT runs its own pilot for aggregated home batteries and devices: premises need 15-minute interval data, the aggregator sends telemetry every two seconds, device-level data must be available on request, and statistical sampling isn’t allowed (ERCOT).
What devices actually deliver
Measured load reduction per device during events
Average kW cut per device in utility evaluations and lab testing.
Swipe to see the full chart →
Two things stand out. Smart thermostats deliver several times more than the old air-conditioner switches they replace, and results vary by brand and climate: the same type of thermostat cut 1.40 kW in Texas but about 0.9 kW in Indiana. CenterPoint measured its thermostats from runtime data supplied by the thermostat platform and converted it to kWh — a device-level method rather than a meter baseline (CenterPoint Energy).
Where estimates and settlements part ways
Pre-cooling
Running the AC harder before an event raises the usage that same-day adjustments look at. For El Paso Electric’s program, Texas set the adjustment window to the two hours starting three hours before the event for this reason (Texas PUC).
Snapback
Devices catch up after an event. Baselines that only measure the event window miss the extra load afterward, which matters for the utility even if it doesn’t change the customer’s pay.
Offline and opted-out devices
Texas evaluations count a thermostat only if it took part in at least half of an event; offline devices and those that drop out sooner are excluded (Texas PUC).
Batteries already cycling
A battery that discharges every evening for time-of-use savings already lowers the meter baseline, so a meter-based event shows less “reduction.” Device-output metering avoids this.
Adjustment caps
Caps such as ±20% stop a strange morning from distorting the baseline but can under-credit real changes on unusually hot days.
Small loads, noisy meters
A 1 kW change is hard to see in one home’s meter data, which is why residential programs lean on control groups, deemed values, and device telemetry.
How VPP Economics will use M&V
Estimates are useful for deciding whether to join. Verification is what decides the check. We plan to bring the grid operators’ own rules into the calculator so you can test an estimate against real data.
Estimates
The Residential VPP Calculator estimates pay from published program terms, with flexible-power defaults taken from the measured results on this page.
Verify mode
Upload interval data from the utility (Green Button) or a device export, pick the grid operator’s baseline rule, and see measured kW per event next to the estimate.
Program view
For program managers: compare deemed and measured performance across a fleet, and the share of enrolled devices that actually respond.
Method and caveats
This page summarizes measurement and verification for nonexperts and isn’t a substitute for a program’s tariff or settlement manual. Baseline rules shown are weekday rules; weekend, holiday, and product-specific rules differ. Checked October 1, 2026.
Measured results come from specific programs, climates, and years and won’t match every home. The roadmap describes planned features, not ones available today.