When the evening peak moves

At noon, an imaginary electricity system has room to supply more demand. At six in the evening, it is working much harder. A planner can respond by adding supply, storage or network capability. Another question comes first: does every activity at six need to happen at six?

Sometimes the answer is yes. Sometimes a vehicle could charge earlier, a process could run in a different interval or thermal storage could separate electricity use from the moment heat is needed. Moving the timetable can provide a service to the power system while preserving the useful activity. Its cost depends on the controls and payments needed to preserve that activity, and on the alternative investment the particular system can avoid.

The proposition becomes more interesting when the timetable belongs to real people. A shift worker, a tenant without controllable equipment and a factory with continuous production do not have the same choices. The technical possibility of moving demand is the start of assessing flexibility, not the amount a system can count on buying.

A small curve with a specific job

Use three one-hour intervals to isolate the arithmetic. The invented system’s demand is 6 megawatts at noon, 10 megawatts at five and 10 megawatts at six. Move one megawatt-hour from each evening interval to noon. Noon rises to 8 megawatts, and the two evening intervals fall to 9 megawatts each.

One-hour interval	Original demand	Demand after the illustrative shift
Noon	6 MW	8 MW
5 p.m.	10 MW	9 MW
6 p.m.	10 MW	9 MW
Energy across these intervals	26 MWh	26 MWh

The selected intervals use the same energy in this simplified example. The peak falls from 10 to 9 megawatts. Nothing here establishes that a real system has spare supply at noon, that every day follows this pattern or that the shift is free of losses. The example only demonstrates the difference between reducing a peak and reducing total energy.

It also defines the job. A resource able to deliver this two-hour shift does not automatically provide flexibility during a different peak, a week of difficult conditions or a rapid stability event. Its availability, notice period and duration have to match the service being purchased. A megawatt without a timetable is an incomplete product.

The IEA’s 2026 Scaling Up Demand Flexibility report examines the role through case studies of South Africa in 2025, Thailand in 2030 and Ireland in 2035. The later cases are forward-looking analyses, not observations of completed future systems. Their different settings are a reminder that flexibility’s value depends on what the power system needs it to do.

Britain’s service now moves in both directions

Britain’s Demand Flexibility Service began as a response to winter pressures, but that origin no longer describes the entire product. NESO’s current service account records its move to year-round operation in November 2024 and the introduction of bidirectional flexibility in April 2026. The page dates the launch of constraint-management procurement capability through DFS to 7 October 2026, replacing procurement of those actions through the Local Constraint Market. That is NESO’s stated service change; it does not report the outcome of every subsequent dispatch.

Bidirectional is the significant word. A useful action can involve increasing demand at one time as well as reducing or moving it at another. That is consistent with a system in which abundant generation and local network conditions can make the difficult hour something other than the familiar evening peak.

The April update also introduced zonal procurement. The location is part of the product: increasing consumption where generation would otherwise be constrained can help use that supply without demanding an equivalent extra transfer through a bottleneck. Increasing consumption elsewhere may not address the same constraint. A meter recording a successful increase in use is therefore only one part of verification; the action must also be in the zone and interval the operator requested.

NESO’s March 2026 announcement explains that suppliers and aggregators connect participating households and businesses to the service. The consumer’s experience therefore depends on an intermediary’s offer, communications and reward structure. A system-level service design is not itself a promise of a particular household payment.

This is where a simple “use less electricity” message becomes inadequate. The system may need a particular action, in a particular place and interval. Uncoordinated shifting can solve one problem while failing to address the constraint that prompted the request. The market needs to describe the service clearly enough that providers can deliver the useful version of flexibility.

The thing being moved is a service

An electric vehicle needs sufficient charge by the time it leaves, not necessarily at the first minute it is plugged in. A building needs appropriate indoor conditions, not necessarily continuous identical compressor operation. An industrial process needs its product and quality, not necessarily an inflexible schedule where a different one is technically acceptable.

These distinctions create opportunities, but the user’s requirement remains the boundary. A vehicle departing unexpectedly, a room with little thermal tolerance or a process with tightly coupled stages can change the available flexibility. A program should learn those constraints rather than treat every connected device as an equally dispatchable resource.

The US Department of Energy’s discussion of grid-interactive buildings describes pre-heating and pre-cooling as ways of moving electricity use through a building’s thermal behavior. It also distinguishes services lasting seconds from those lasting hours. The discussion is an explanation of possibilities, not proof that any particular home can shift its load comfortably for an arbitrary period.

Automation can help by acting within agreed limits without requiring a person to watch prices. But “automated” does not mean unconstrained. The controls need usable settings, reliable communications and an intelligible override. The household or operator should know what is controlled and what happens when its plans change.

There is a difference between permission to shift a charge and permission to leave someone without the service they need. A design that counts on participants not noticing or not understanding the control is unlikely to sustain trust. The useful resource is a well-specified area of flexibility, not the largest amount a device can technically be prevented from consuming.

The rebound has to fit somewhere

Load moved out of an event often returns elsewhere. In the invented curve, that was the point: two megawatt-hours moved to noon. In a real program, the location of the returning demand matters. If many devices resume simultaneously after an event, the new pattern can create a different peak.

A system assessment should therefore examine the surrounding hours, not crop its chart to the rewarded interval. It should distinguish energy that was shifted from energy not used at all and include relevant losses or changes in operation. A short favorable segment cannot establish the full result.

Different devices also respond differently over time. A delayed process may need a catch-up period. Thermal operation may change with weather and occupancy. A charging session can be constrained by connection time and required departure energy. A provider’s portfolio needs a model of these patterns that is tested against operation, rather than assume every dispatchable kilowatt is interchangeable.

Coordination can make the rebound manageable, but it has a cost and a design. Signals, controls and contractual arrangements should support the useful aggregate pattern while respecting participants’ limits. The performance measure needs to reflect the delivered service, including what happens after the event ends.

The baseline is an estimate of a missing day

To reward a reduction, a program often needs an estimate of what demand would have been without the event. That baseline is a counterfactual. It cannot be read directly from the same meter at the same time, because the household experienced only one version of the day.

This creates a measurement problem even when everybody acts honestly. Weather, occupancy and changing routines can alter expected consumption. A baseline suited to one load pattern may perform less well for another. The method needs to be appropriate to the service and transparent enough for participants to understand the basis of a reward or dispute.

It also affects system planning. Enrollment is not delivered capacity. A nominated capability is not necessarily the realized response, and a realized response on a mild day may not establish availability during the system’s hardest interval. Evidence from relevant events and conditions matters before flexibility is used to defer another investment.

That does not require the resource to be perfectly predictable; conventional equipment is not perfectly available either. It requires comparable treatment of uncertainty. The question is what dependable contribution the portfolio provides under the conditions in which it is needed, and what other arrangements cover the remaining risk.

Flexibility has a distribution

A household already minimizing electricity use may have little discretionary load left to move. A household with an electric vehicle, controllable heating and spare financial capacity may have more opportunities. Rewarding measured flexibility can be reasonable, but it should not be mistaken for rewarding equal effort or equal need.

The Centre for Sustainable Energy’s Access to Flexibility research examines low-income and vulnerable households’ participation in National Grid DSO flexibility markets. Its project account, updated in February 2025, identifies disparities in service provision and the need to investigate barriers. Its scope is distribution-level markets, not a measured verdict on every national program or every household’s willingness to participate.

The policy question is how to broaden useful access without inventing flexibility that a household does not have. Assistance with equipment, better intermediaries or simpler enrollment may help in some settings. Other households may need protection from unfavorable tariffs rather than a promise that they can earn enough from shifting.

Essential energy use also needs to remain distinct from avoidable inconvenience. A program should not rely on people accepting conditions that conflict with their needs to obtain a reward. The user’s service boundary and opt-out must remain meaningful when the system is stressed, not only when participation is easy.

The controls need a relationship with the user

Participation also creates information flows. A provider may need meter or device data to establish availability and measure delivery. The customer needs to understand which information is used, who receives it and what continues after leaving. A service can require appropriate verification without treating access to every connected device as the default price of participation.

The arrangement must survive ordinary changes: a replacement charger, a new tenant, a failed connection or someone taking responsibility for a relative’s account. These are not exotic edge cases in a service intended to operate for years. Clear records and a workable support route keep the agreed limits attached to the right customer and equipment. Otherwise, a control strategy that performs well in a trial can become confusing as the installed fleet changes around it.

When the timetable is cheaper, count why

The business case includes controls, metering, communications, aggregation, payments and participant effort. It also includes the alternative actually avoided or deferred. A reduction at the wrong location may have little value for a local network constraint even while reducing the system-wide peak.

Cost comparisons should therefore identify the service, place and period. They should not take the price of a short demand-response event and compare it with an entire power station as though both provide the same capability. Nor should a possible future deferral be booked as a saving before its reliability and timing have been established.

Flexibility remains a powerful idea because it begins with a useful question about demand rather than assuming supply must follow every existing habit. Its best version preserves people’s services while changing the electricity schedule. The curve improves because a workable timetable has been agreed, measured and paid for—not because a planner found a way to make the people disappear from it.