Don’t take our word for it.

Transcribed from a published report

The figure is copied as printed from the cited page of the source document. We do not adjust it.

1. Download the source document from the URL in the evidence registry.
2. Go to the cited page and table.
3. Copy the figure and its units, period and any footnotes exactly.
4. A second reviewer re-reads the page before the figure is marked verified.

Limitations

• We rely on the publisher's own definitions, which can change between years.
• Restated prior-year figures are recorded against the report that states them.

Executive remuneration: single-figure method

We use the audited single-figure table in NESO's Annual Report on Remuneration, broken into components. Payments the report mentions but leaves out of its total are shown separately and never folded in.

1. Read the single-figure table (salary, benefits, pension, annual award, long-term award, total).
2. Record each component with its basis: paid in year, awarded for the year, vested in year, or employer contribution.
3. Check that the included components add up to the reported total.
4. Record payments in the footnotes (e.g. retention payments) as separate components, flagged as outside the total.
5. Record the contractual salary rate separately from salary actually paid in the year.

Limitations

• The single figure includes long-term awards in the year they vest, which may relate to performance in earlier years.
• Pension is shown as the employer contribution, not the value to the individual.
• Outstanding unvested awards are not valued in the report and are not shown as amounts here.

Annual constraint costs from NESO daily data

We add up NESO's daily constraint costs for each financial year from the Constraint Breakdown dataset, one CSV per year.

1. Download the financial-year CSV from the NESO Data Portal.
2. Check that every row falls inside the expected financial year.
3. Add up the 'Thermal constraints cost' column for thermal costs.
4. For total constraint costs, add the thermal, voltage, inertia and largest-loss cost columns.
5. For volume, add up 'Thermal constraints volume' and report the absolute value in TWh.

Reproduce with scripts/aggregate-neso.mjs

Limitations

• Tags can be revised after the event; we report the dataset as downloaded on the accessed date.
• Days missing from the published file are not filled in.
• Constraint costs are a subset of total balancing costs. Do not add them to the balancing cost figure.

Day-ahead demand forecast error

We take a simple mean of NESO's published half-hourly absolute percentage error (APE) between the day-ahead national demand forecast and outturn, for each financial year.

1. Download the Day Ahead Half Hourly Demand Forecast Performance CSV.
2. Assign each half-hour to a financial year (1 April – 31 March) by its settlement date.
3. Average the APE column across all half-hours with a numeric value.
4. Flag incomplete years and leave them out of trend comparisons.

Reproduce with scripts/aggregate-neso.mjs

Limitations

• A simple mean weights every half-hour equally, regardless of demand level.
• NESO's incentive scheme metric may apply exclusions or different weightings.
• Embedded solar growth makes national demand harder to forecast over time, so a rising error is not purely a measure of effort.

Annual national demand energy

We convert half-hourly national demand (average MW) from NESO's Historic Demand Data to energy (MWh) by multiplying by 0.5 hours, then add up the results for each financial year.

1. Use the ND column of NESO's Historic Demand Data (calendar-year files combined into financial years).
2. Multiply each half-hourly MW value by 0.5 to get MWh.
3. Add up the values for each financial year and convert to TWh.

Reproduce with scripts/aggregate-neso.mjs

Limitations

• National demand leaves out embedded generation, exports and pumping, so it is smaller than total consumption.
• The dataset is published about 21 days in arrears and NESO makes retrospective corrections.

Balancing cost per MWh of national demand

Annual system balancing costs from the statutory accounts, divided by annual national demand energy calculated from NESO data.

1. Take system balancing costs (£m) for the financial year from the annual report.
2. Take national demand energy (TWh) for the same financial year from our calculation above.
3. Divide £m by TWh. This gives £ per MWh directly, because £1m / 1 TWh = £1/MWh.

Limitations

• The numerator and denominator come from different sources with different definitions.
• Balancing costs are recovered through BSUoS charges on a different basis, so this is not the BSUoS tariff.
• Using a smaller demand base (national demand rather than total consumption) gives a higher £/MWh.

Electricity Margin Notice episodes

We count EMN episodes in Elexon's SYSWARN feed. An episode is a sequence of EMN issue and update messages closed by a cancellation message.

1. Query the SYSWARN API one month at a time and keep the raw JSON.
2. Keep messages with warningType 'ELECTRICITY MARGIN NOTICE'.
3. Sort by publish time. Start an episode at the first message and close it at the next cancellation.
4. Assign each episode to the financial year of its first message.

Reproduce with scripts/count-emn.mjs

Limitations

• The API returned nothing before 9 May 2023, so earlier years are not reported.
• Capacity Market Notices and other warnings are a different, lower threshold and are not counted.
• An EMN asks the market for more margin. It is not a blackout or a demand-control event.

Transcribed from an official data workbook or statistical table

The figure is read by the ingest script from the named sheet of the publisher's spreadsheet, and the validator re-checks the stored value against the parsed sheet.

1. npm run data:ingest downloads the workbook, stores it under its content hash and parses the named sheet into data/processed.
2. The verified record states the value and a JSON pointer into the processed file.
3. npm run data:validate recomputes the value from the processed file and fails if they differ.

Reproduce with scripts/ingest/parsers/balancing.ts, scripts/ingest/parsers/consumer.ts

Limitations

• Publishers restate earlier years in later editions; the site uses the latest edition and keeps earlier ones in data/processed.

Operational balancing costs from NESO's Annual Balancing Cost Report

The sum of the cost categories in the workbook's Total outturn sheet for each financial year.

1. Parse the Total outturn sheet: constraints, energy imbalance, minor components, reactive, reserve, response and restoration (£).
2. Add the categories. The absolute volume column is excluded.
3. Cross-check FY2025-26 against the £3.1bn stated in the outturn report.

Reproduce with scripts/ingest/parsers/balancing.ts

Limitations

• Not the same as the statutory accounts figure. Both are kept and labelled.
• Energy imbalance can be negative, so the total is net.

Wind curtailment share

Curtailed wind divided by NESO's 'unconstrained wind outturn' (outturn plus curtailment), from the same workbook.

1. Read curtailment (GWh) and unconstrained outturn (GWh).
2. Divide and multiply by 100.

Reproduce with scripts/ingest/parsers/balancing.ts

Limitations

• Uses NESO's definition of unconstrained output.

Wind turn-down reconstructed from NESO Wind BOA Volumes

Sum of negative bid-offer acceptance volumes on wind BM units for each financial year. Used to cross-check NESO's published curtailment.

1. Download each financial-year file from the NESO Data Portal (Wind BOA Volumes).
2. Sum BOA_Volume where negative (accepted bids that reduce output) and report the absolute value.
3. Compare with the curtailment figure in NESO's Annual Balancing Cost Report workbook.

Reproduce with scripts/ingest/parsers/neso-portal.ts

Limitations

• Covers only BMUs in NESO's incentive wind forecast set.
• Provisional until reviewed.

Low-margin hours from de-rated margin forecasts

Counts settlement periods where NESO's 4-hour-ahead de-rated margin forecast (Elexon LOLPDRM) is below a threshold, using the latest forecast published at that horizon.

1. Query Elexon's loss-of-load and de-rated margin forecast in 7-day windows.
2. For each settlement period and horizon, keep the most recently published forecast.
3. Count periods below 500, 1,000 and 2,000 MW at the 4-hour horizon, and divide by two to get hours.

Reproduce with scripts/ingest/parsers/elexon.ts

Limitations

• Thresholds are our choice, not NESO's operational thresholds.
• A forecast, not an outturn: tight forecasts usually trigger market responses.