How balancing offers are screened, period by period

Day-ahead surveillance reconstructs a merit order, because the order book is not published, and every finding there carries the reconstruction's own error. Here the regulatory feed publishes the individual balancing energy bids, so this view reads the ladder rather than estimating it. There is no model error to disclaim.

What the publication does not carry is identity. It can show that one price step carried the volume that set the price; it can never show whose step it was. Every finding attributes to a step in a ladder.

Every quarter-hour of the delivery day is screened in both directions, whichever one was actually called. The inputs are already in memory and the screens are arithmetic, so the whole day costs no extra upstream call - and a control that chose one direction would only hide the other.

The ladder

Every offer published for the quarter-hour, cheapest first, per balancing product. Products are called in sequence and each keeps its own price axis.

The trailing band

The distribution of that same quarter-hour on preceding delivery days. A screen fires when a reading clears that band, so a volatile period needs a larger move than a quiet one - a period is measured against its own history rather than a fixed constant.

What is not on the ladder

Not everything that balances the zone comes off it. Activation by component splits what actually answered, and a product that supplied nothing has no settled price to be measured against, so its price-relative readings are absent rather than zero.

Nothing crosses on the ladder charts, and that is correct rather than a rendering gap: only one direction is called in a quarter-hour, and the reference lines are the settled marginal prices rather than a point where supply met demand.

Findings for the selected quarter-hour appear directly below the day strip, with upward and downward readings side by side. A direction with too little trailing history says Not screened. The detailed ladder and threshold tables are folded below each direction; open them to inspect the near misses.

What answered

The activation split by component. The products are called in sequence, and not all of what balances the zone comes off this ladder at all.

Product ladder details

Offered, unavailable and priced, per balancing product. A product that supplied nothing has no settled price to be measured against, so its price-relative readings are absent rather than zero.

What fired

The findings for this direction and quarter-hour, each stating what was observed with the numbers in it.

Threshold readings

Every screen listed whether or not it fired, including the ones this zone's publisher cannot support. A screen that is silently absent is indistinguishable from one that found nothing, so none of them is ever omitted - the reading says which it is.

The screens, and what each is looking for. None of them claims abuse. In Belgium every reading is about volume of a given shape and never about whose, because the ladder is published without participants in it. In Great Britain the units are named, which makes the last screen below possible and makes restraint necessary: a named unit appears in a finding's evidence, where a reviewer can check it, and never in its heading.

Steep ladder where it was called

The euro move per 100 MW into the product's own ladder at the volume actually activated. A steep ladder at the called volume is where a small withheld block moves the settled price a long way, so it is context that raises the value of the other two.

One offer at the called price

Price-invariant volume sitting at exactly the price the quarter-hour settled at: one offer carried that much MW and chose the price. The aggregation cannot say whose offer it was.

Less volume offered than usual

Volume offered into this direction fell against the distribution of that same quarter-hour on preceding delivery days. A period is measured against its own history, so a routinely thin quarter-hour does not fire and a routinely deep one that thins does.

One unit set almost all of an unusual price

Great Britain only, because only its publisher says which accepted actions set the cash-out price and with what weight. It fires when a single named unit carried 90% or more of that weight and the price itself was at least two sigma from the median for the same settlement period across preceding delivery days. Both conditions are needed: one unit setting the whole price is ordinary in GB, and on 35 of 80 measured periods that is exactly what happened.

Depth price is read 100 MW into each product's own ladder, not 100 MW along the shared volume axis. The products are called in sequence and the second one starts a few hundred MW in, so an absolute depth would land before its first offer and describe a product that had not begun. At or better is defined only for the product the settled price is attributed to, because it is a comparison against a price the other products were not paid.

The strip is every quarter-hour of the delivery day, both directions screened, so the day can be scanned before any one period is opened. Selecting a period changes the ladder below and rides in the URL.

The compact price chart shows the day's imbalance prices and marks the selected quarter-hour. Open Settlement replay for the price components and minute-by-minute system conditions, or Day-ahead vs imbalance to compare the same period's two prices. These links keep the date and exact UTC period. The folded settlement ledger includes every quarter-hour, with N/A for missing readings, and exports as CSV with the time of the read.

Belgium and Great Britain, and the two are not the same object.

Belgium: individual bids from the market platform change of 2024-05-22. Delivery days before that were published under a schema this does not read, which is not a gap in collection - nothing here ever covered that schema. The operator publishes the ranked ladder itself, so there is no model between it and the screen.

Great Britain: the operator ranks nothing. Each unit publishes its bid and offer prices, its physical notification and its export and import limits, and the merit order here is built from them: a unit's offerable volume is the band between what it said it would run at and the most it could, cut into the price steps it declared. That is a reconstruction and it carries those assumptions, which is why the page header says so before any number. What GB adds in return is identity: the units are named, and the publisher states which accepted actions set the cash-out price and with what weight. Re-deriving the price from those weights matched the published price on 77 of 80 settlement periods measured on 2026-10-05, to within one penny per MWh.

A GB period is measured once the publisher's next settlement run has revised it, which is the morning after delivery. Before that the period is shown but not screened, and the strip says so rather than reading as a quiet market. That run is also not the final reconciliation: later settlement runs can still move a number, and nothing here re-measures them.

Not every GB price is set inside the mechanism. On 12% of the periods measured the cash-out price was answered entirely by services the operator buys under their own contracts rather than by bids into the mechanism, and the page names those: the service, the party and the asset, from the published adjustment feed. They carry no ladder, so there is nothing to compare their price against and no screen reads them. Knowing a price came from there rather than from the merit order is what changes how the period should be read.

Frequency response is a different market. Dynamic Containment, Moderation and Regulation are bought in separate response auctions and paid for being available, so they produce no offer in the mechanism and no action in the settlement stack, and no screen on this page reads them. For Great Britain one observation relates the two, and it has its own section below.

Three kinds of GB headroom are kept apart, because conflating the last two would manufacture a finding. Volume priced where somebody could take it is offered. Volume priced at the publisher's refusal price was declared and put out of reach. Volume with no published price at all was never offered, and a unit that chose not to offer and a unit that published nothing look identical from here - 324 of 826 units with headroom published no prices for it in the period measured.

The publisher holds a rolling two years; a delivery day past that is no longer served and nothing here stored a copy. An empty response is an unconfirmed read: it does not establish that no bids were published. The default view moves back to the most recent day the feed returns, saying so and carrying that day in the link you share. Partial days name how many quarter-hours have a returned ladder.

A unit can hold a contracted frequency response obligation and also offer or bid in the balancing mechanism. The operator's published guidance says such a unit should protect the response it has sold: for the low-frequency obligation by lowering its export limit so no accepted offer can take that volume, for the high-frequency one by raising its import limit, or by pricing the volume out of reach. This section reads, for the selected settlement period and each side, whether any priced volume was left inside the range the obligation reserves.

It measures against that guidance, which says "should". The binding service terms were not read, so a unit outside the guidance has not been shown to break a rule. "Priced out" means a price of GBP 9,999/MWh or more, the level the guidance's own example uses and the level this product already treats as a refusal; volume priced high but below it is counted as exposed, which overstates exposure for a unit that prices out at, say, GBP 3,000.

It shows the units holding an obligation and the obligation in MW, how many could be assessed, how many were exposed and the MW exposed with the largest exposed unit, how many were not assessable, and how many had no published position. The line above it counts, across the day, the measured periods with any exposed unit on either side.

It is an observation, not a finding. Exposure is not an acceptance, and an acceptance is not non-delivery: nothing here says a unit failed to deliver. The capacity used is a proxy taken from the mechanism's unit register, not the register response is contracted against, and the two disagree: aggregated units are often registered at 0 MW, and some units hold more obligation than their registered capacity. A unit whose capacity is missing, zero, smaller than its own obligation or contradicted by its own limit is counted as not assessable, never as exposed. A unit with no position in the mechanism at all, because it is not a balancing mechanism unit, is counted separately as having no published position. Both counts are the observation's own error and are shown beside its figures, including the day line. Naming the unit with the most exposed volume is naming what was published, not attributing conduct to it. Nothing is scored or thresholded.

Three empty states differ. A period with no row has not been measured yet. A period whose figures are blank was measured before this observation existed, or the obligation could not be read. A period with zero units had no obligation on that side.

The bids are as published, so the ladder is not in doubt. The inference from it is. A unit already at minimum stable load, a network constraint, an outage not yet disclosed, a position already hedged: each produces the shapes these screens look for, and none is visible in this data. Verify against outage disclosures before drawing any inference.