Methodology

This page documents how Gauge Intelligence computes every figure in the public archive. The methodology is versioned, reproducible, and never paywalled. When a calculation method changes, the revision is dated and documented here. For the release schedule and the distinction between revisions and corrections of error, see Release cadence.

Arrival-to-fifteen punctuality (A2F)

Gauge Intelligence measures freight punctuality using an all-cause, fifteen-minute threshold. A journey segment is classified as punctual if its delay at the measurement point is fifteen minutes or less. The percentage of punctual segments over a given period is the A2F punctuality figure.

This is an all-cause measure. It captures delay from every source: infrastructure faults, operator-caused incidents, weather, possessions, and third-party events. A2F therefore differs from Network Rail’s Freight Cancellations and Lateness (FCaL) metric, which captures only Network-Rail-attributed events.

Gauge Intelligence’s A2F will therefore report lower punctuality than Network Rail’s FCaL for the same services and time window. FCaL excludes operator-caused delay; A2F includes it. Both measures are valid: A2F answers “how reliably did freight arrive?” FCaL answers “how much of the delay was the infrastructure manager’s responsibility?”

Delay is computed by comparing the TRUST actual arrival timestamp against the CIF scheduled arrival time at each measurement point. A journey segment records delay_minutes (the timetable variation reported by TRUST at the corridor exit point). Segments where delay_minutes is nil are treated as zero delay (on time).

The fifteen-minute threshold is the industry standard for freight punctuality, consistent with ORR’s published freight performance framework.

Data sources

Gauge Intelligence processes the following data sources. All processing is programmatic; no manual adjustments are applied to performance figures.

NROD / TRUST — Network Rail’s real-time train movement feed, ingested continuously. Every train activation, movement, cancellation, reinstatement, and change of identity on the national network is received and processed. This is the primary source for journey-level timing, delay, and cancellation data. Refresh cadence: continuous (real-time streaming).

CIF timetable — the Common Interface File provides scheduled timings against which actual performance is measured. Each journey’s planned arrival is derived from CIF timing points at the terminus location. Refresh cadence: periodic (updated when Network Rail publishes revised schedules).

BPLAN geography — Network Rail’s track topology dataset, mapping STANOX location codes to physical routes. Used to assign journey segments to infrastructure corridors. Refresh cadence: weekly (sourced from the Rail Data Marketplace).

Historic Delay Attribution (HDA) — Network Rail’s periodic publication of DAPR cause codes, assigning each delay event to a responsible party and cause category. Used for delay reason breakdowns and Schedule 8 attribution direction (whether Network Rail or the operator caused the delay). Refresh cadence: periodic download.

ORR published data — the Office of Rail and Road’s national freight statistics, including Table 1325 (freight moved by commodity and route). Used for regulatory context and cross-referencing against official aggregate figures. Refresh cadence: annual.

For how these feeds are connected, parsed, and corrected before any figure is computed — the STOMP consumer, TRUST message processing, and the timestamp and schedule-type corrections applied at ingest — see Data ingestion.

Detailed data source documentation is available to licensed users, including ingestion timestamps, record volumes, and freshness monitoring.

Operator league tables

Operator performance rankings are published as both aggregate and corridor-stratified views. Aggregate figures can reverse when operators serve different mixes of corridors (a form of Simpson’s paradox). For the methodology behind operator rankings and Simpson’s paradox detection, see Operator league tables.

For the methodology behind anomaly detection (statistical process control, change-point detection, and resistant summary statistics), see Anomaly detection. For causal inference: how Gauge Intelligence separates association from causation in published claims, including Bradford Hill scoring and DAG prerequisites, see Causal inference. For the four categories of rail freight activity that report in aggregate on the supply-chain page rather than as per-site register entries — Network Rail yards, nuclear logistics, defence logistics, and infrastructure-project flows — see Aggregate coverage.

Known limitations

For a complete treatment of TRUST observability boundaries (which services are invisible to the data feed and why), see Data window: what TRUST tells us, and what it cannot.

Daylight saving time correction. TRUST timestamps contain a known bug during British Summer Time: certain timestamp fields are offset by one hour (3,600,000 milliseconds). Gauge Intelligence applies a systematic correction to affected fields (gbtt_timestamp, planned_timestamp, actual_timestamp, canx_timestamp, dep_timestamp, orig_dep_timestamp, original_loc_timestamp) during BST periods. This correction is applied at ingest, before any performance calculation.

Deduced activations. Some train movements arrive without a matching TRUST activation message. This can occur when the activation message is lost in transit, when a train is activated outside the normal messaging window, or during system outages. When movements are received for an unactivated train, Gauge Intelligence infers the activation from the available movement data to ensure the journey is not lost from the record.

Chain of Identity tracking. A freight train’s identity can change mid-journey, for example when a train is re-signalled or combined with another service. TRUST records these as Change of Identity (CoI) messages. Gauge Intelligence follows the full chain of identities (up to ten links) to reconstruct the complete journey from origin to destination, rather than treating each identity as a separate service.

Attribution coverage. Delay cause codes from the Historic Delay Attribution dataset are not yet fully categorised for all delay minutes. Where attribution data is unavailable for a specific journey, delay is recorded but not assigned to a cause category. Coverage is improving as successive HDA releases are ingested.

Terminal arrival source. Not all journeys record an arrival at the scheduled destination. When a destination arrival is not observed, the system falls back to the last terminated movement, then to the last recorded movement. Delay figures derived from fallback sources are excluded from punctuality calculations to avoid misleading intermediate-point comparisons.

Schedule 8 estimates

Gauge Intelligence publishes estimates of bilateral Schedule 8 payment obligations between Network Rail and freight operating companies. These are estimates, not confirmed settlement figures. Actual settlement amounts are commercially confidential between Network Rail and each operator.

The estimates use published Control Period 7 (CP7) payment rates. CP7 base rates are set at 2022-23 prices and indexed annually to CPI (see derivation below):

• NR-to-FOC direction: £25.81 per minute (2022-23 base); £28.42 per minute in 2025-26 after CPI indexation. This is the rate Network Rail pays the operator when delay is caused by Network Rail.
• FOC-to-NR direction: £51.98 per minute (2022-23 base); £57.24 per minute in 2025-26 after CPI indexation. This is the rate the operator pays Network Rail when delay is caused by the operator.

The two-times differential means that operators pay Network Rail roughly twice per minute of delay what Network Rail pays operators. This asymmetry is a deliberate feature of the CP7 Schedule 8 regime, reflecting the operator’s bearing of greater commercial risk for delay it causes.

Schedule 8 CPI indexation

Schedule 8 rates are indexed annually to CPI under Schedule 7, para 2.7.2(a) of the freight track access contracts:

Indexed rate = Base rate × (1 + (CPI(t-1) − CPI(2022)) / CPI(2022))

Where CPI(2022) is the annual average for calendar year 2022 (the contract base) and CPI(t-1) is the annual average for the calendar year prior to the rate applying. The index used is ONS series D7BT (CPI All Items, 2015 = 100, annual average).

Rates apply per the operator lens: Schedule 8 settles bilaterally between Network Rail and each FOC. See league-table methodology for lens reconciliation.

Calendar year	Annual average CPI (D7BT)	Used for rate year
2022	121.66	base — no rate applies
2023	130.54	2024-25
2024	133.98	2025-26
2025	138.68	2026-27

2025-26 derivation (uses calendar-year-2024 CPI):

• Indexation factor = 1 + (133.98 − 121.66) / 121.66 = 1.101240
• NR-to-FOC = 2581p × 1.101240 = 2842.30p → 2842p (£28.42/min)
• FOC-to-NR = 5198p × 1.101240 = 5724.24p → 5724p (£57.24/min)

For comparison, 2024-25 used the calendar-year-2023 CPI average. Indexed rates were £27.69 per minute (2024-25 CPI-indexed from £25.81 base) paid by Network Rail to operators and £55.78 per minute (2024-25 CPI-indexed from £51.98 base) paid by operators to Network Rail.

The application source-of-truth for indexed rates is the schedule8_payment_rates table (seeded with the values above). Customer-facing £/min citations across this archive resolve to the row covering the journey’s run date.

Exposure estimates are computed per journey: delay minutes above a benchmark threshold are multiplied by the applicable rate. The benchmark threshold is scaled by journey distance (per 100 train operator miles), consistent with the Schedule 8 contract structure.

Sub-threshold analysis (quantifying the delay minutes that fall below Schedule 8 reporting thresholds and are therefore invisible to the settlement process) is available to licensed users.

Four analytical lenses

The same underlying journey data can be viewed through four analytical lenses. These are query dimensions, not domain boundaries. A single journey appears in all four lenses simultaneously.

Operator. Performance grouped by freight operating company (FOC). Identifies which operators are experiencing the most delay and Schedule 8 exposure on a given corridor or nationally.

Destination. Performance grouped by terminal, port, or yard. Identifies which end points are accumulating delay, regardless of which operator or corridor serves them.

Corridor. Performance grouped by infrastructure route. A corridor is defined as an ordered sequence of STANOX locations along a physical rail route. Journey segments are assigned to corridors by matching TRUST movement locations against corridor point definitions. A single journey may traverse multiple corridors. For how corridors are delineated — Gauge Intelligence adopts Network Rail’s Freight Network Study corridor set rather than drawing its own — see Corridor selection.

Supply chain. End-to-end performance across connected services and routes. Captures the cumulative effect of delay as it propagates through intermodal supply chains from port to distribution hub.

Corridor-level and operator-level analysis for specific periods is published in the public archive. Deeper cross-dimensional analysis (for example, operator performance on a specific corridor, or destination performance filtered by supply chain) is available under commercial licence.

Period-ahead forecasts

All period-ahead punctuality forecasts use a seasonal-naive baseline: the same period from the prior financial year. No forecast is published without a 90% prediction interval. Where no prior-year baseline exists (inaugural year), no forecast is published and the absence is disclosed.

For full methodology, see Forecasting methodology: seasonal-naive baseline.

For delay attribution methodology (why DAPR cause-code shares are compositional data and how Gauge Intelligence compares attribution distributions across periods), see Delay attribution. For corridor recovery time after disruption, see Recovery time. For chart design methodology, see Visualisation.

Rail periods

All published figures are aligned to Network Rail’s thirteen-period financial year. Each period is approximately 28 days. Periods are referenced using the notation P[number] [financial year], for example Period 13 2025-26.

This alignment ensures that Gauge Intelligence figures are directly comparable with Network Rail’s own periodic reporting and with ORR’s regulatory publications.

Version history

Version 1.1 — April 2026. Added cross-references to five new methodology pages: anomaly detection, causal inference, delay attribution, recovery time, and visualisation.

Version 1.0 — April 2026. Initial publication covering A2F punctuality, Schedule 8 exposure estimation, four-lens analytics, and data source documentation. Applies to all figures published from the GEML Period 13 2025-26 report onward.

Methodology changes will be dated and documented in this section. When a methodology change affects previously published figures, a correction notice will be posted on the corrections page with full provenance.

Note on rail-year notation: ORR’s CP7 PR23 source documents render rail years in slash form (year over two-digit successor); this archive normalises to hyphenated form (e.g. 2022-23) for consistency across publications.