Flexibility Markets

Implicit vs. explicit flexibility

Implicit flexibility responds to prices — a battery arbitraging the intraday spread, a consumer on a dynamic tariff shifting load. Explicit flexibility is sold as a product to a system operator. Good design makes the two complementary; bad design pays explicitly for behaviour the price signal would have produced for free.

The buyer landscape

The same megawatt of flexibility can serve the energy markets, the TSO's balancing needs, the TSO's or DSO's congestion needs, and adequacy mechanisms. Each buyer runs different products, timelines and prequalification — the value-stacking problem is really a market-fragmentation problem.

Local flexibility markets

Platforms such as GOPACS in the Netherlands, Piclo in Great Britain and NODES in the Nordics let network operators procure congestion relief from distributed assets at specific grid locations. Location is the product: a megawatt behind the right substation is worth what the reinforcement it defers would have cost.

A decade of pilots has converged on one product architecture, documented across seven countries by the EU DSO Entity's 2026 survey: a capacity leg procured in advance (€/MW availability, pay-as-bid, tendered from seasons to years ahead) whose winning offers flow automatically into an energy leg (€/MWh utilisation) activated day-ahead to intraday, open also to non-reserved bids. The architectural philosophies behind the platforms differ more than the products: GOPACS is a coordination layer over the existing power exchanges — congestion bids clear through EPEX and ETPA order books, each matched with a counter-bid outside the congested area so redispatch never unbalances the system; NODES and Piclo are independent marketplaces; Rome's RomeFlex is a DSO-built end-to-end stack down to a certified device at the customer's premises. Scale is no longer trivial: UK Power Networks passed 300,000 registered assets in June 2026 (see the use case below), and Areti's 22 MW from 1,300 Roman customers anchors a plan to meet a 3.3 GW peak with 2.6 GW of wires — cutting total expenditure by nearly half against pure reinforcement.

Congestion management vs. balancing

Balancing is about system-wide energy equilibrium; congestion management is about where power flows. The distinction matters because a single zonal price cannot see location — which is why congestion flexibility is procured through separate, locational instruments layered on top of the zonal market.

Storage and demand response as providers

Batteries, electrolysers, heat pumps, EV fleets and industrial processes each bring different activation speed, duration and predictability. Product definitions written for generators — symmetric, always-available, hours-long — systematically undervalue them; the reform agenda is largely about rewriting those defaults.

The EU demand-response framework

The forthcoming European network code on demand response extends harmonised market access rules for aggregation, demand response and storage down into distribution networks — the first EU-level attempt to standardise how local flexibility is bought.

The code standardises the vocabulary the pilots invented: controllable units, service-providing units and groups as the resource hierarchy; a four-stage qualification ladder (provider qualification, product prequalification, ex-post verification, grid prequalification by every affected operator); and flexibility registers as the data backbone — Enedis' register already holds over 700,000 flexible sites and doubles yearly. The decisive layer, though, is national: the code delegates how flexibility is actually accessed, procured, activated and validated to National Terms and Conditions, which is where the seven pilot playbooks will either harden into rules or be flattened into lowest-common-denominator compromise. That fight — not the code itself — will determine whether Europe gets one flexibility market design or twenty-seven.

What an FSP actually is

A flexibility service provider is the party that qualifies, bids, receives the activation instruction, is settled and is penalised. It may be an independent aggregator, a supplier, an energy-service company or the asset owner acting alone; at transmission level the same role is called a balancing service provider, and in most markets the same company holds both. The EU demand-response network code fixes the hierarchy underneath: controllable units (the physical devices) are grouped into service-providing units and groups, and it is the group, not the device, that is prequalified and dispatched. Article 17 of the Electricity Directive lets an independent aggregator act without the customer's supplier consenting — the legal move that turned FSPs into a market segment rather than a supplier add-on.

What the FSP sells is not the flexibility of any one asset but the reliability it manufactures from many. A single heat pump is unavailable whenever its owner objects, a single EV whenever it is unplugged; five thousand of either behave like a dispatchable plant with a known availability curve. Portfolio diversity converts behavioural noise into a firm product, and the FSP's margin is the difference between what the portfolio can be contracted for and what any member could promise alone. That is why the provisions that decide an FSP's economics are the ones about pooling: whether assets can be substituted within a portfolio, whether a portfolio may span network zones, and whether the baseline is set per device or per group.

Reading a resource

Five parameters describe any flexible resource well enough to see which products it can serve: the power it can change, the duration it can hold that change, its response speed, the rebound it produces afterwards (the energy it must recover once released), and its predictability — how far availability depends on weather, occupants or a plug being in a socket. Generator-shaped products assume large, symmetric, hours-long, instantly available and always-there; almost nothing on the demand side is all five at once, and each departure from the template is a design decision waiting to be made.

Batteries

Grid-scale batteries are the one demand-side resource that matches the generator template on four of the five parameters: symmetric, sub-second, predictable and, within a portfolio, always available. The exception is duration, and it is the exception that shapes every battery rule. A one-hour battery bidding four hours of reserve is a fiction, so products either cap the duration they demand (Great Britain's Dynamic Containment was designed around a 15-minute delivery envelope precisely so that batteries could hold it) or impose state-of-charge management rules that force the provider to keep energy in hand and recover it between activations. The German FCR requirement that an energy-limited unit be able to sustain full activation for a set period, and recharge in a way that does not itself disturb the system, is the archetype: a product rule that exists only because the resource is a battery.

Batteries are also the first resource to have saturated a product. Frequency-containment prices in Great Britain and Germany fell steeply as battery capacity arrived, and revenue migrated to wholesale arbitrage, to the slower reserves and, increasingly, to congestion. That migration is the argument for locational value: a battery on the right medium-voltage feeder is worth more to the DSO than a larger one at transmission, and it is also the risk, because a battery chasing the wholesale price charges when solar is abundant, which is precisely when the feeder it sits on is most stressed. The connection queue, not the market, is now the binding constraint on how many are built (see Networks).

Electric vehicles

The flexible resource is not the car but the charging session, defined by three numbers: when it was plugged in, when it must leave, and how much energy it needs in between. Smart charging shifts that energy within the window; vehicle-to-grid adds a discharge, bounded by the battery and by whatever floor the owner sets. Fleets at depots are contractually firm and behave like small industrial loads; private cars are behaviour-driven and firm only in aggregate, with high availability overnight and very little in the working day. EVs bite on the low-voltage network, where a street of evening charging coincides with the residential peak — the congestion that drove the Dutch and British LV programmes described below.

The design difficulty is that a charging session has no observable counterfactual. A car that charged at 02:00 instead of 18:00 consumed the same energy; the only baseline is a model of what it would otherwise have done, and that model can be gamed by anyone who controls the plug-in time. Explicit ex-post products therefore treat EVs badly, and the designs that work are the ones that dispense with the baseline: dynamic or time-of-use tariffs and capacity-based network charges that make the shift worth doing without measuring it (the implicit route), scheduled products that contract a charging plan in advance rather than a deviation from one, or direct controllability with a regulated compensation, as in Germany's rule that lets a DSO dim controllable loads to a guaranteed minimum in return for reduced network charges. Smart chargepoint regulations that default charging to off-peak hours, as in Great Britain, are the same idea done at the appliance standard rather than the market.

Heat pumps and thermal resources

A heat pump's flexibility is the building's thermal mass and its hot-water cylinder: typically thirty minutes to two hours of turn-down before comfort binds, less in a poorly insulated house, more with a buffer tank. Two features distinguish it from every other resource. The rebound is real and often larger than the reduction, because the building must be reheated and a heat pump run harder does so at a lower coefficient of performance. And availability is inversely correlated with need: the coldest hours are the ones the DSO most wants relief in and the ones in which the heat pump can least afford to stop. Products built for heat pumps therefore need weather-conditional availability, short activations with a stated rebound window, and verification that does not rest on a noisy ex-post baseline — which is why certified controllability and regulated dimming, rather than markets, are the instruments most countries have reached for at the mass-market end.

Thermal storage changes the picture entirely. Electric boilers and large heat pumps feeding a district-heating store, or a hot-water tank in a home, turn heat into hours or days of storage with almost no rebound, and they are predominantly a turn-up resource: they absorb surplus power in negative-price hours and provide downward balancing. Power-to-heat in Denmark, Germany and the Netherlands is already among the largest sources of downward flexibility on the system, and the district-heating combination of a CHP plant, an electric boiler and a store is a two-directional resource that can either generate or consume on demand. Industrial thermal buffers — cold stores, furnaces, drying processes — sit in the same class: predictable, long, and worth more to the energy market than to any congestion product.

Industrial and commercial load

Interruptible industry was the original demand response. Aluminium smelters, chlor-alkali plants, cement mills, paper machines and cold storage offer large blocks that respond within seconds or minutes, contracted firmly, and they have supplied interruptibility schemes and slow reserves for decades. Their limit is the recovery: a process can usually be curtailed once a day and must then run to catch up, so the resource is a single shot rather than a repeatable one, and it is most valuable in the scarcity hours when the product is least likely to be needed twice. Electrolysers are the coming addition and the most ambiguous: physically they can move in both directions faster than any thermal plant, but their own economics want a high load factor, so the flexibility is available only if the market pays more than the hydrogen it displaces. Commercial buildings are the smaller cousin of the heat pump — HVAC turn-down for thirty to sixty minutes with a rebound — and data centres, with their uninterruptible power supplies and shiftable workloads, are the resource everyone is now trying to reach and almost no one has yet contracted.

What follows for product design

Because every product parameter screens resources, a product is a choice of providers whether or not the buyer intended one. A one-megawatt minimum bid admits only aggregators and industry; a symmetry requirement excludes EV charging, which can only turn down; a four-hour duration excludes heat pumps and most commercial buildings; an availability window in the working day excludes homes; an ex-post baseline penalises every behaviour-driven resource in favour of batteries, which have a clean one. Reform of demand-side products is largely the reversal of these defaults, one at a time: asymmetric products, shorter durations, kilowatt-scale minimums, portfolio baselines, scheduled and baseline-free alternatives.

The other lesson is about where the boundary between explicit and implicit should fall. For batteries and industrial load, explicit products work: the resource is firm, metered and contractable. For millions of EVs and heat pumps, the transaction costs of qualification, baselining and settlement are out of proportion to the value of each device, and the effective instruments are the ones that shape behaviour without contracting it — tariffs, network charges, appliance standards and regulated controllability — with an aggregator layered on top only where a pool is large enough to be sold as a firm product. The demand-side design question is not which market the heat pump should bid into; it is whether it should be bidding at all.

Why coordination is hard

A distribution-connected battery activated upward by the TSO for balancing can overload the very feeder the DSO needed relieved. Without coordination, each operator's remedy is the other's disturbance — the physics ignores the institutional boundary.

Coordination models

Designs range from TSO-led (the DSO merely validates), through market-based platforms with joint procurement, to traffic-light schemes where DSO network state constrains TSO access to distributed bids. The chosen model allocates not just activation priority but market access itself.

The operating models in production span a full spectrum. The Netherlands runs an integrated platform where every congestion trade is matched with a counter-trade to stay balance-neutral. France draws the line physically: RTE sets operational limits at each transmission-distribution transformer and Enedis optimises freely below them — a clean interface targeted for full deployment by 2028. Norway sequences access: DSOs activate first and the TSO takes the residue. Italy shares one marketplace — Areti procures on the same GME platform as Terna, with DSO priority and differentiated pricing to prevent double payment. Each model answers the same question — who may use a distributed asset, when, without destabilising the other operator's grid — and none has yet proven dominant.

Baselines

Explicit demand-response payment requires knowing what consumption would have been — an unobservable counterfactual. Baseline methodologies decide the product's integrity: generous baselines invite phantom flexibility; harsh ones punish genuine response. No design question in flexibility markets is more underestimated.

The method families in use range from declarative baselines the provider nominates itself, through historical averages with same-day adjustment (eight-of-ten-days rules), regressions, and zero baselines for storage, to ex-post techniques — metering before and after, control groups, rebound-catching windows. Practice is converging pragmatically: simple standard profiles for small domestic assets, nominated baselines with statistical accuracy checks for industrial ones. The subtler integrity problem is the compensation effect — a site's battery discharges as contracted while its heating system, behind the same connection, quietly ramps up: measured at the device the service was delivered, measured at the grid it never happened. Validating delivery at the connection point against everything behind it, using smart-meter data, is where the current regulatory argument sits.

The methods actually in use across European markets sort into a short list: historical averages or medians over the last X days of the same type; the British “mean X-in-Y”, which averages X of the last Y days after discarding the highest and lowest; nearest-neighbour selection of the most similar recent days, and comparison against a control group of similar units that did not respond, both used by Enedis; simple averages of the hours immediately before activation, which suit consumption sites best; zero baselines; and provider-nominated forecasts, subject to validation. Several operators bolt a correction factor onto the historical figure, derived from the site’s measured profile in the couple of hours before activation — precisely because historical methods break when the weather moves. The reviewed verdict is that no single method fits every technology or aggregation level, and the more pointed critique is that baseline-settled services suit local markets badly whenever an aggregator is continuously optimising the same assets across several markets at once; capacity-limitation products — paid for staying under an agreed ceiling — are the proposed alternative. The Dutch answer, settling against a revised commercial schedule instead of an estimated counterfactual, is the same instinct implemented differently.

Prequalification cascades

An aggregated portfolio may need product prequalification from the TSO, grid prequalification from several DSOs, and registration in a flexibility register. Every additional gate raises the minimum viable portfolio size — process design is entry-barrier design.

Data and observability

DSOs historically ran networks with little real-time visibility. Flexibility procurement forces investment in metering, forecasting and flexibility registers — and raises governance questions about who may see, and monetise, distributed asset data.

Settlement chains

When an aggregator activates a customer's flexibility, the customer's supplier sees an imbalance it did not cause. Transfer-of-energy rules and perimeter corrections between aggregator, supplier and BRP are the plumbing that makes independent aggregation financially fair — and their absence is why it stalls.

Market-based or cost-based congestion relief?

Market-based redispatch lets flexibility set its price but invites the inc-dec game — assets positioning in the wholesale market to be paid for relieving congestion they helped cause. Cost-based redispatch (the German Redispatch 2.0 choice) prevents the game but extinguishes the price signal. This is the central unresolved argument in European congestion design.

Locational granularity

Flexibility value is locational; zonal markets are not. Options range from node-specific tenders through constraint-zone products to network-tariff signals. Finer granularity prices reality better but fragments liquidity into markets sometimes one substation deep — where a single provider is a monopolist.

Availability vs. activation payment

Paying for availability buys certainty and favours capital-intensive assets; paying only on activation shifts risk to providers and favours opportunistic ones. Most real schemes blend both — the ratio is effectively a choice of which provider type the market should attract.

Firmness and penalties

A flexibility product is worth what happens when it fails to deliver. Penalty design must be strong enough that operators can rely on the product instead of holding redundant conventional remedies — but penalties calibrated for power plants can be existential for a small aggregator.

Practice has resolved this more gently than theory predicted: nearly every operating market uses delivery tolerances instead of penalties. Payment is pro-rata above a minimum delivery threshold — commonly 60–85% of the contracted volume — and capped somewhat above 100%; miss the floor and you simply are not paid. Only the Dutch market imposes true penalties. This is a deliberate infant-market choice: DSOs judge that liquidity is currently scarcer than firmness, and that punitive regimes would empty order books they spent years filling. The penalties will come when the markets can afford them — and delivery rates around 75% in the mature British market suggest the tolerance era has real costs too.

Procurement timeframe

Long-term contracts underwrite investment in flexibility that does not yet exist; short-term procurement prices the actual system state. Mature designs run both: framework agreements for capability, near-real-time markets for dispatch.

Interaction with network tariffs

Time-varying and capacity-based network tariffs are themselves flexibility instruments. Procuring flexibility a well-designed tariff would have elicited anyway double-pays; the coherent design treats tariffs, connection agreements and flexibility markets as one toolkit.

Grids that overbook

The frontier application is not deferring reinforcement but designing networks that assume flexibility from day one. Dutch DSOs may administratively overbook substations to 110–150% of nominal capacity, managing everything above the physical rating with contracted flexibility under a regulated cost cap; Norwegian and Swedish operators run 20–30% overbooking. Enedis goes furthest, embedding flexibility into the connection design of primary substations — its pilot added 210 MW of renewable connection capacity on ten substations for roughly 30% less capital, at a curtailment cost below 0.06% of energy. The honest caveat from every case study: reinforcement still often beats even free flexibility in the cost-benefit test, and connection acceleration — not deferral — is emerging as flexibility's most valuable product in queue-choked grids.

The liquidity problem

Every operating market names thin participation as its binding risk, and the remedies converged too: publish procurement horizons years ahead so investment can respond, push entry barriers down (minimum bids have fallen from 1 MW toward 100 kW, and as low as 1 kW in Norway and 300 W in Rome), publish every auction result so pay-as-bid markets still generate price discovery, and defer punitive obligations until the order book can bear them. Behind all of it sits a regulatory-economics problem: an operator rewarded on capital expenditure has no reason to prefer flexibility over copper. The British answer — a total-expenditure regime letting the network keep half of what flexibility saves — is the clearest demonstration that provider liquidity follows buyer incentives, not the other way round.

One directive, many answers

The 2019 Electricity Directive told member states to let distribution operators buy flexibility through transparent, non-discriminatory, market-based mechanisms — technology-neutral, open to aggregation, and integrated with the markets the TSO runs — unless the national regulator finds that a market would not be economically efficient and grants an exemption. What emerged is not one market but a spread. Great Britain moved first and furthest: all six distribution operators signed the Energy Networks Association's Flexibility Commitment in December 2018 and have procured against a common product portfolio ever since. The Netherlands launched GOPACS the same year as a shared TSO–DSO congestion platform. France followed in 2020 with Enedis's zonal auctions and Sweden with the Sthlmflex pilot; Portugal opened FIRMe in 2023, Slovenia a low-voltage scheme the same year, and Italy three separate pilots in 2023–24, each designed by a different DSO under one regulatory resolution.

The British portfolio is the reference design

Five standardised active-power services, defined once by the industry body and procured individually by each operator. Peak Reduction and Scheduled Utilisation are bought seasons or months ahead against predefined settlement periods, need no dispatch signal at all — delivery is simply expected in the stated window — and are paid only for energy delivered. Operational Utilisation is dispatched against real network events, with a two-minute or fifteen-minute activation deadline, or instructed a week ahead; again paid only on use. The last two, Scheduled and Variable Availability with Operational Utilisation, reserve capacity months ahead and pay both an availability rate and a utilisation rate, the variable version allowing the reserved quantity to be adjusted closer to delivery. All carry a thirty-minute minimum delivery, and the provider — not the network operator — chooses which of its qualified assets inside the relevant grid area actually responds.

The portfolio's value is that it separates the two things a network operator buys: the option, and the energy. A constraint the operator can forecast a season ahead needs no option at all, so it pays only for delivery; a constraint that might or might not bite needs the option, and pays for it. Most emerging markets offer one product and discover this distinction the hard way.

GOPACS solves the baseline problem by not having one

The Dutch platform procures intraday redispatch, contracted close to real time against requests from the TSO or a DSO, and day-ahead capacity restriction paid per MW of agreed reduction. Its deeper design choice is in settlement. Because every accepted bid is a change to a market party's commercial schedule, and each congestion trade is matched by an opposing trade outside the constrained area so that relieving congestion never creates an imbalance, delivery is verified by comparing the revised schedule against metered outturn. No counterfactual has to be estimated — the baseline problem below simply does not arise. The price of that elegance is that participation runs through wholesale order books, which sets a floor under who can realistically play.

What the flexibility is worth, and what the prices actually show

The economic ceiling on an availability payment is the network investment it defers, and Italy is one of the few places where that logic is written into the rules: E-distribuzione caps the availability price at the avoided grid development cost — what the fit-and-forget reinforcement would have cost — while RomeFlex and MiNDFlex cap availability at €30,000/MW/year and utilisation at €500/MWh. Observed prices, though, say more about liquidity than about scarcity. In the British 2023–24 round one operator's thin auctions accepted availability bids around €11,000/MW/h and utilisation around €7,000/MWh, orders of magnitude above its five peers, because too few providers were bidding to discipline anyone. E-distribuzione's first Italian auctions cleared availability at a weighted €863/MW/h with utilisation close to the €500 cap across a handful of activations; Slovenia's low-voltage scheme paid about €0.60/kWh. Contracted volumes tell the same story from the other side: Great Britain and the Netherlands are an order of magnitude ahead of everyone else, and the rest is still pilot scale.

The negative result worth keeping

Sweden's Sthlmflex, launched in 2020 to relieve winter congestion around Stockholm, has been wound up. Liquidity never reached the level at which a market does anything a bilateral contract could not, and Ellevio concluded that local flexibility markets are not, in the near term, the right instrument for Stockholm. That is the most useful single data point on this page. A local flexibility market is not free: it costs design, platform, prequalification and settlement effort on both sides, and it repays that only where congestion is deep and recurrent enough, and providers numerous enough, for competition to be real. Where those conditions are absent, the honest answer is a connection agreement, a tariff, or copper.

What the practitioners say

Sixteen structured interviews — six network operators, seven service providers, three platform operators — rank the barriers differently by role, and the disagreement is itself the finding. Providers name economic viability first: activation counts are unpredictable, local needs are published with little notice, and the revenue will not justify buying control equipment, so participation falls back on conventional commercial and industrial units that an operator can adjust by hand. Their asks are concrete — availability payments even at the expense of utilisation payments, and three-to-five-year published outlooks of where flexibility will be needed. Network operators name liquidity first, and trace it to two things they own: perimeters drawn tightly enough to fix the constraint are often too small to contain enough providers, and distribution networks remain thinly observed, so activations rest on static models and historical data and are still issued by telephone, e-mail and SMS. Platform operators report that their systems could carry far more volume than they are given. All three groups name regulatory fragmentation. The common thread is not that the market design is wrong in the abstract: it is that the market is too small, too late-notified and too manual to pay for the equipment that would make it work — which is why the forthcoming Network Code on Demand Response, with common terms for aggregation, baselining, flexibility registers and TSO–DSO coordination, matters more than any single national design.

What is actually running

UK Power Networks distributes electricity to London, the South East and the East of England — three licence areas, roughly a third of Britain's customers. It appointed EPEX SPOT as its market platform provider in November 2023 after a competitive tender, opened day-ahead flexibility trading on the Localflex platform in April 2024 and long-term reservation the following month. By January 2025 the platform carried over 100,000 live assets, representing more than 2 GW available for activation; by June 2026 it had passed 300,000. The commercial premise is stated plainly in the price control: a flexibility-first approach worth a reduction of up to £410 million in load-related expenditure across the current regulatory period. Every design choice below follows from that premise — the operator is not running a market for its own sake, it is buying an alternative to copper and has to prove the alternative is cheaper.

The areas: a zone can be one street transformer

A flexibility zone is a network constraint with a boundary drawn around it, published per tender round. The high-voltage zones are the familiar kind — a grid or primary substation whose firm capacity is forecast to run out. The low-voltage tier is the newer and more radical part: the zone is a single secondary transformer, which is to say a few streets. The operator's own open data makes the scale legible. Across tender rounds nine to fourteen, 1,793 low-voltage zone entries have been published covering 852 distinct secondary transformers, of which 312 are live in the current round; high-voltage zones add several hundred more. The distribution is lopsided in a way that maps onto the network rather than onto policy: the East of England accounts for roughly three-fifths of the low-voltage entries, London and the South East for the rest.

Granularity is the central trade-off and this is where to see it. A zone drawn tightly enough to fix the constraint is a zone in which very few assets happen to sit, which is exactly the liquidity problem network operators name first in the interviews above. Publishing postcode-level data for each competition is the mitigation: a provider can check whether it is even in the right place before spending anything on qualification.

Three products, three commitments

The portfolio replaced the earlier Sustain, Secure and Dynamic labels with three services that differ in what the provider is actually promising. Long-Term Scheduled Utilisation commits a provider to reduce demand inside pre-agreed windows, on contracts running up to two and a half years, paid only for energy delivered. Long-Term Scheduled Availability with Operational Utilisation reserves the capacity for defined windows and confirms the day before whether it will be used, paying both for standing ready and for delivering. Day-Ahead Scheduled Utilisation, live since April 2024, asks for no commitment at all beyond tomorrow: the provider enters a daily auction and prices its energy afresh each day.

The participation terms are as important as the products. The minimum is 10 kW of flexible capacity in a zone, single or aggregated, with a thirty-minute minimum run; a household below that threshold participates through a third party; individual assets and aggregations both bid as a single Flexible Unit; and assets that do not yet exist can bid into long-term tenders, which is how a market procures investment rather than merely dispatching what is already there. Observed utilisation prices have reached around £600/MWh, varying by location — the value of relieving a constraint is a local number, not a national one.

The process, end to end

The chain begins years before any bid. Distribution Future Energy Scenarios locate where and when capacity runs short; the Distribution Network Options Assessment then decides between reinforcement and flexibility using the industry's Common Evaluation Methodology, which also fixes the budget and the guide price for each zone — the buyer's willingness to pay is derived from the network investment being deferred, not negotiated. The long-term tender then runs in four stages: needs and guide prices published openly with free registration; prequalification, where the most common failure is simply that the asset is connected to the wrong part of the network; competition, where bids are assessed on value, volume and budget, with availability and utilisation fees reduced to a single comparable rate in £/MWh so that structurally different offers can be ranked against each other; and award on the standard industry contract, followed by a proving test before the service counts as deliverable. Results are then published in granular form — bids by area, volumes accepted and rejected, provider names, contracted volumes, and the availability and utilisation prices received.

Day to day the market runs on a fixed clock: requirements for the following day are published by 10:00 and utilisation decisions confirmed by 13:30, with providers free to reprice daily. Dispatch is a merit order with three filters. Is the need identified at procurement still valid? Do primacy rules disqualify a unit because it has been sold to the system operator for something else — the discipline that stops the same megawatt being promised twice? And which remaining unit is cheapest in total, which is not the same as cheapest per megawatt-hour: a generator with a low £/MWh price but a long minimum run time can cost more overall than a dearer unit that stops when the constraint does. It is the clearest illustration on this page of why a flexibility product is a bundle of technical parameters and not just a price.

Registered, contracted, dispatched — three different numbers

The headline figure is participation, and it is genuinely large. The operating figures are smaller and more instructive. In its first year the day-ahead market dispatched 4.4 GWh across more than forty zones and more than 150 competitions; the published dispatch record has grown to some fifty thousand entries by September 2026. Set that beside a fleet of 300,000 registered assets and the arithmetic is stark: a few tens of kilowatt-hours per asset per year.

That is not a failure, but it must be read correctly. Congestion relief is an option, not an energy business: the network needs the asset to be there in the handful of hours when the constraint binds, and the rest of the year it should be doing something else — which is precisely why availability payments exist and why providers, in the interviews, ask for more of their revenue to come from them. The three numbers should therefore be kept apart in any claim about a flexibility market's success. Registration measures reach. Contracted capacity measures how much the network operator was willing to buy. Dispatched energy measures how often the constraint actually bound. A market can be healthy with a low third number and a high second; it cannot be healthy with a high first and nothing behind it.

What travels, and what does not

Two things in this case are portable. The first is transparency treated as market infrastructure rather than as a reporting obligation: needs, guide prices, postcodes, baseline datasets for electric vehicles, heat pumps and households, tender results with names and prices, and a public dispatch log. A 10 kW asset cannot afford due diligence, so the operator publishes enough that due diligence is unnecessary. The second is the platform itself, and its lineage is worth noticing: the same exchange built the ENERA pilot in Germany, connected its trading system to GOPACS in the Netherlands in 2023, then built the auction-based platform used here — and has since been selected by RTE and Enedis for a French local flexibility pilot. The British market is being exported as a template.

What does not travel is the premise. This works because congestion is deep and spreading across three licence areas, because a price control put a number on the capital expenditure that flexibility must beat, and because a single large operator could afford to build the market and publish everything about it. Where a network faces sporadic congestion or has few operators to serve, the same design would cost more than it returns — which is where the diagnosis below comes in.

Read the system before designing the market

Dronne, Roques and Saguan's comparison of the German, Dutch, British and French initiatives sets out the diagnostic that the map above otherwise lacks. Four features of the local system, they argue, explain the design choices better than any first-principles reasoning about market efficiency.

The type and frequency of congestion. Injection congestion — too much local generation — is mostly a question of optimising the curtailment of assets that already exist. Load congestion, driven by EV charging and heat pumps, usually calls for flexibility that has yet to be built. And depth matters before either: heavy, recurrent congestion can repay the cost of standing up a market, sporadic congestion cannot.

Whether the flexibility already exists. Germany and the Netherlands began with substantial untapped potential — the paper's illustration is a single eastern German operator whose 16,000 night-storage water heaters represent 370 MW against a 1,200 MW peak. Great Britain's purpose was closer to the opposite: to bring new resources into being, against an estimate that flexibility could avoid £4–13 billion of distribution investment by 2050. France concluded that roughly half of its future needs could be met by resources already installed.

But potential is not reach. The distinction the framework leaves implicit has since become the binding one: flexibility counts only if it can be metered, addressed and settled. Germany is the case in point, and not in the direction its resource base suggests. Intelligent metering systems covered around 3% of German metering points in mid-2025 — even against the statutory mandatory-rollout cases the quota stood near a sixth, and the regulator opened proceedings against dozens of default metering operators in March 2026 — while the legal requirement to reach 90% of all metering points does not bite until 2032. France, by contrast, finished its Linky rollout at roughly 95% coverage, and one British operator has 300,000 assets registered on a single platform. A megawatt behind an unmetered, uncontrolled connection is not a market resource but an aspiration, which is precisely what network operators mean when they name observability as their binding constraint.

How many operators share the problem. Germany has four TSOs and some 880 distribution operators, several working the same territory at different voltage levels; France has one TSO and one distribution operator serving 95% of consumers, with the Netherlands and Great Britain in between. The count is not trivia. It determines how much coordination the system needs, and coordination between many parties is expensive enough that a shared external platform can beat every operator building its own market and negotiating bilaterally with the rest.

What the incumbent mechanism already does. The alternatives to a market are reinforcement, differentiated network tariffs, capped or "smart" connection contracts, limits on traded volumes, and redispatch — which is either cost-based, paying audited costs, or market-based, paying bids. France already redispatched on a merit order through the balancing mechanism, but with a 10 MW minimum bid and without publishing activated offers: economic, yet opaque and closed to small assets. Germany relied on network solutions and feed-in management, curtailing renewables above 100 kW at 95% cost reimbursement with no merit order between offers. Great Britain leaned on network solutions and accelerated "connect and manage" connections. The Netherlands had no market-based distribution mechanism at all. A market adds different value in each case: a price signal where there was none, transparency where there was a closed one, or access where the incumbent excluded small resources.

Four diagnoses, four designs

Crossing the local situation with the design choices gives the mapping the paper is named for. Where congestion is already significant and the flexibility exists, the gain is dispatch efficiency, and the design is a transparent short-term merit order — ENERA's day-ahead and intraday product on the exchange. Where congestion is significant but the resources are not there, the gain is investment, and the design is long-term: capacity reservation alongside activation payment, multi-year contracts, published needs — UKPN contracting up to seven years and shortlisting assets that do not yet exist. Where many operators share the same resources, the gain is coordination, and the design is an external platform that centralises needs and screens activations so that one operator's relief is not another's constraint — the GOPACS case. And where the incumbent mechanism works but excludes, the gain is access, and the design lowers thresholds: 10 kW at low voltage in Britain against the 10 MW minimum of the French balancing mechanism, no exclusivity so a resource can serve several markets, and penalties proportionate enough that a newcomer can bear the risk.

What follows from it

There is no one-size-fits-all congestion market, and the authors doubt a uniform European design will emerge — though common principles may. Five years on, the evidence both confirms and qualifies that. The plumbing has converged: nearly every market now runs pay-as-bid, most pair an availability leg with a utilisation leg, and thresholds are falling everywhere. What has not converged is what the market is for, and that is exactly what the four questions predict. Britain, facing both kinds of congestion and needing new resources, built the long-horizon, capacity-paying, low-threshold portfolio. The Netherlands, with many operators and flexibility already connected, built the shared short-term platform. France, with light congestion and resources in place, built low-barrier zonal tenders and left the heavy lifting to its existing mechanisms. Germany is the instructive exception, and not because its potential is small. Deep injection congestion, some 880 network operators, and a distribution system it largely cannot see: with intelligent metering on a few per cent of metering points there was little for a local market to clear against, and the inc–dec exposure of one would have been real. So Germany extended cost-based redispatch to distribution-connected units of 100 kW and above instead — an administrative answer to a problem whose market answer was not yet available to it. Four systems, four answers, each legible from the diagnosis rather than from the theory — and a reminder that metering and observability are not implementation details downstream of market design. They decide whether there can be a market at all.