Demand response for commercial accounts

Demand response is a capacity product, not a cheaper rate

A supply contract prices the energy a site consumes. A demand charge prices the peak the site sets on the utility's system. Demand response prices something else again: the availability of a reduction at a moment the grid operator chooses. The three sit on the same meter and answer to different rules, which is why a site can carry a well-priced supply contract, an expensive demand charge and a demand response enrolment that pays it for capability nobody called on. The payment is for being available inside a defined window, not for the energy saved, and the energy saved is not the test of whether the commitment was met. What is tested is the difference between the load the site would have carried and the load it actually carried, interval by interval, during the event. Nearly everything difficult about demand response follows from that sentence.

The contract is with an intermediary, never the grid operator

In every market the site is not the market participant. ERCOT pays the qualified scheduling entity that submitted the winning offer, not the building. PJM pays its members, which it calls curtailment service providers, and those providers pay the end-use customer under a private contract. ISO New England settles with a qualified market participant that holds the registered resource. CAISO requires a demand response provider agreement and a scheduling coordinator before a load can bid at all. The programme rules therefore set the ceiling on what is possible, while the intermediary's contract sets what the site actually receives, what it must prove and who carries the risk of a shortfall.

• The committed quantity in megawatts, and whether it varies by season or time of day
• The baseline method, who calculates it, and what happens when the site's normal load shifts
• Notification window, maximum event duration, and the number of events the product permits
• Test obligations, including unannounced tests, and who is present for them
• What a shortfall costs: a reduced payment, a de-rating for future periods, or a charge
• Whether the site can decline an individual event, and what declining costs

ERCOT: emergency response is bought by term and paid for availability

ERCOT's Emergency Response Service is procured rather than subscribed to. ERCOT selects offers submitted in response to a request for proposal and pays the winning qualified scheduling entity a capacity payment for making resources available for deployment. The trigger is specific: ERS may be deployed as an operational tool when physical responsive capability falls below 3,000 MW and is not projected to recover above 3,000 MW within thirty minutes of Non-Spin deployment, and ERCOT may also deploy it at EEA Level 3. Two products exist by ramp rate — ERS-10 must complete its ramp within ten minutes and ERS-30 within thirty. The term is short and seasonal. The technical requirements for the summer 2026 term run from 1 June to 30 September 2026, and the preceding winter term ran from 1 December 2025 to 31 March 2026, with its resettlement still being posted in October 2026. Offers are made per standing Time Period, with a megawatt quantity and a declared maximum base load that may differ between periods. After a release instruction the resource has ten hours to return to a condition that lets it meet its obligation again — the clause operations managers underestimate, because the curtailment ends when ERCOT says so and the recovery is the site's problem.

PJM and ISO New England: the commitment lives in the capacity market

In PJM, demand response is a voluntary programme in which retail customers are paid for reducing load when PJM requests it during high prices or a threat to reliability, with payments flowing through curtailment service providers. The quantity a site helps clear is bought in the Reliability Pricing Model, whose base residual auction is held roughly three years before the June-to-May delivery year; just over 8,000 MW of demand response was offered into the 2026/27 auction on an installed-capacity basis, essentially unchanged from the year before. In ISO New England the equivalent route is the Forward Capacity Market, where a demand resource is qualified for a commitment period and must maintain that qualified capacity, respond within its registered parameters and meet the metering and verification requirements, with financial consequences if it does not. The difference from ERCOT's seasonal service is the horizon. A multi-year capacity commitment is made against a site whose production plan, occupancy or equipment may not survive the term, so assignment, exit and change-of-quantity terms deserve more scrutiny than the headline quantity. A capacity obligation is also not a promise to shut the building down: it is a promise to deliver a verified reduction, which is why the next section is the one that decides whether the promise is kept.

The baseline is the product, and it decides the payment

Performance is calculated as a baseline minus what the meter recorded. The baseline estimates what the site would have consumed in that interval had there been no event, so the payment rests on a counterfactual. ERCOT publishes seven default baseline types in its Demand Response Baseline Methodologies document; the qualified scheduling entity selects one from those ERCOT offers, that selection then applies across every Time Period and Contract Period of the term, and some types need at least twelve months of interval history before ERCOT can model the load at all. Under the maximum-base-load method the availability factor is computed from the average fifteen-minute load during the committed Time Period minus the declared maximum base load. A baseline is a description of a building's normal week, and it goes stale. A site that adds a line, loses a tenant or changes shift is measured against a past that no longer exists, and the error runs in whichever direction hurts. Do the arithmetic on your own interval data before signing: take the ten most comparable weekdays, average the load in the hours you would be called, subtract the load you can credibly shed in those hours without breaching a process, a temperature or a compliance limit, then repeat for a hot week and a mild one. The smaller result is the quantity worth offering. The larger one is how a payment reduction happens.

Metering, tests and the paperwork that has to arrive on time

A reduction is worth what the meter data can prove. CAISO requires ISO-certified revenue-quality metering and, for ancillary services, direct telemetry, so a load cannot simply report its own curtailment. ERCOT requires each submetered premise inside an ERS resource to be represented by a single unique meter identifier, and the qualified scheduling entity must submit meter data within thirty-five days of a deployment or an unannounced test; missing or invalid data can be treated as a failure to meet the performance requirement whatever the site actually did. Provisional substitute resources must pass an unannounced test before they count. The organisational version of the same trap is that the person who can curtail the load is rarely the person who holds the metering relationship, and the thirty-five-day clock does not care. Name an owner for meter data before enrolment rather than after the first event, and check that the utility or transmission distributor will release interval data for the specific meter identifiers in the resource.

Notice, ramp, and the operating limit that is not negotiable

Notice windows differ by product and are the first thing to hold against the site's process. A ten or thirty minute ramp is compatible with switching off chillers and raising a setpoint. It is not compatible with a batch that must run to completion, a cold chain that cannot drift, or a clean room whose pressure cascade cannot be interrupted. PJM publishes a thirty-minute notification exception guideline precisely because some loads cannot move faster, which is an argument for matching the product to the process instead of promising a number the operations team cannot deliver. Recovery is the other half of the same question. A resource released from an event has a fixed period to return to a state where it can meet its obligation again, and a site that has spent its thermal storage, chilled water or compressed-air reserve may need longer than the product allows. Curtailment capability is therefore not one number: it is a set of numbers by season, by hour and by duration, and the smallest of them is the one to commit.

What to send, and what this page does not do

This page explains the mechanism; it enrols nobody. The commitment itself is contracted with a curtailment service provider, a qualified scheduling entity or a demand response provider, and it is a separate contract from the supply agreement, frequently with a different counterparty. What a supply review contributes is the load analysis the rest of the decision depends on: the interval history, the load factor, the peak that sets the demand charge and the blocks that could be shed without harming the operation. A business that does not know its own load shape cannot size a commitment honestly, and an oversized commitment is worse than none because the shortfall is measured against the site's own baseline.

• Twelve months of interval data per meter, or the longest history the utility will release
• A site list with service addresses, account numbers and meter identifiers
• The load blocks that can be shed, by season and by hour, with the duration each can hold
• The process constraint: what cannot be interrupted, and how long recovery takes afterwards
• Whether there is behind-the-meter generation, its nameplate rating and how it is dispatched
• Who operates the building day to day, and who is authorised to commit it

PUT THE OPTIONS SIDE BY SIDE

Your comparison checklist

Common questions