Aggregate coverage — why some categories report in aggregate rather than per-site

Most of the Gauge Intelligence archive resolves to a per-entity register entry. A freight operating company has an operator page. A maritime gateway has a port page. An intermodal hub has an inland-terminal page. An industrial end consumer has a supply-chain page. The operator ran the train, the port loaded the box, the terminal accepted the swap, the steelworks consumed the slab — each one is a recognisable counterparty in the rail freight value chain, and each one earns its own register page with its own carrier mix and its own dated performance figures.

Four categories of activity do not resolve cleanly to that model. They appear on the freight network, they consume a measurable share of network capacity, and they are recognisable to anyone working in the industry — but the attribution model that lets us publish a per-site performance figure either does not fit or does not yet exist. Rather than invent a register entry the data cannot support, Gauge Intelligence lists those categories in aggregate on the supply-chain page and explains, here, what each one is and why it sits in the aggregated block. This page is the standing methodology reference for the four categories.

Network Rail yards and junctions

A Network Rail yard is operational infrastructure. A train enters the yard, swaps a locomotive or a crew, waits for a path, and departs to a destination operated by somebody else. The yard is the verb in the sentence; the destination is the noun. The operational performance that matters to a shipper or a regulator — was the freight delivered on time, was the contract honoured, did the operator meet its Schedule 8 benchmark — attributes to the operator running the service and the destination terminal accepting the delivery. It does not attribute to the yard, because the yard is not accountable for whether the train continued on time or arrived at its booked slot.

Publishing per-yard performance figures would imply otherwise. A “Wembley Yard on-time arrival rate” reads like a statement about Wembley, when the underlying data is a statement about the operators whose trains happen to pass through Wembley on the way somewhere else. The figure would mix operators, traffic types, time-of-day patterns, and arrival routings into a single number that nobody owns and nobody can act on. That is the failure mode the per-site register is designed to avoid.

The yards listed on the supply-chain page span a wide range of activity in the rolling ninety-day TRUST data window: the larger marshalling yards land in the high hundreds to low thousands of endpoints, the smaller pass-through loops in the tens. Wembley, Eastleigh, Toton, Hoo Junction, Carlisle Kingmoor, Whitemoor, Bescot, Tonbridge West, Tees, Tyne, and Acton are the named marshalling yards and junctions where TRUST records the origin or destination of a freight movement. Where the yard is the recorded terminus — typically because the train ends its scheduled diagram at the reception sidings before being re-allocated — the journey counts above. Where the yard is mid-route only, the journey does not appear in the yard’s count; it attributes to whichever destination the operator booked. Crewe Basford Hall, Ipswich, and Colchester Goods Loop sit under the same yard taxonomy and appear in the same table. Doncaster’s pass-through sidings (Belmont Down Yard, Up Decoy and the Wood Yard CE sidings) carry the same shape but remain on the existing Doncaster Yard inland-terminal register entry for historical reasons.

Nuclear logistics

Nuclear flask movements between Sellafield, Heysham, Hartlepool, Hinkley, Hunterston, Torness, and Wylfa are operated principally by Direct Rail Services under long-running contracts with the Nuclear Decommissioning Authority and the generators. The traffic is real, the safety record is the strongest of any rail freight category, and the operational discipline involved is well-documented in the industry.

The choice to report nuclear logistics in aggregate rather than per-site is not a data-availability constraint. The flask movements are visible in TRUST. The choice is a security-disclosure question: the per-site cadence of flask movements between named generation and reprocessing sites is the kind of information that a public archive should not publish without first consulting the operators, the Office for Nuclear Regulation, and the wider sector on what level of detail is acceptable in the public register and what should sit behind commercial-licence access only. That consultation is open. Until it closes, the supply-chain page lists nuclear logistics as a single aggregate category. The per-site detail is available under commercial licence on the terms set out on the licensing page.

Defence logistics

Ministry of Defence freight terminals — Bicester, Marchwood, Catterick, Kineton — handle military stores, ammunition, and vehicle movements. The sites are public knowledge and the train operators are visible in TRUST; the operational sensitivity attaches to the cadence and routing of named flows rather than to the existence of the terminals.

The defence-logistics category sits in aggregate for the same reason as nuclear logistics: per-site cadence is the part that requires sector consultation before publishing in the public register, and the consultation is not closed. The aggregate national figure lands on the supply-chain page; the per-site detail does not.

Infrastructure-project flows

HS2 construction, Sizewell C, Hinkley Point C, and the next generation of major rail-served infrastructure projects move spoil, aggregate, and construction materials through the rail network in flows that look nothing like the steady industrial-site traffic that anchors the rest of the supply-chain register. HS2 spoil moves out through Wembley and Acton on project-specific operator paths but without a project-specific TIPLOC at either end — the recorded origin is a yard, the recorded destination is a disposal site or another yard, and the project identity sits in the operator’s diagram rather than in the freight-service record. Sizewell C and Hinkley Point C have, in the current data window, essentially no rail movements at all; both projects are road-served at this stage of construction.

The right attribution model for project flows is keyed on train UID and operator diagram, not on TIPLOC. The Gauge Intelligence data model is TIPLOC-centric — it is what the four-lens architecture has been built on, and it is what the per-entity register publishes against. A project-lens attribution model that overlays train-UID and operator-diagram information on top of the TIPLOC-centric record is under consideration for a later edition, and the criteria for adding it sit on the release-cadence page. Until then, the project flows that do run are visible through the operator and corridor registers, and the supply-chain page records the absence of a discrete project-lens entry as honest scoping rather than as a feature gap to be glossed over.

Why the aggregated block sits alongside the per-site block

The supply-chain page lists the eight industrial-site pilots above the aggregated block deliberately. The pilots — Drax, Margam, Scunthorpe, Westbury, the Mendip quarries, Boulby, Peak Forest, and Earles — are the shape the per-site register is designed for: a single end consumer, a recognisable carrier mix, a stable TIPLOC footprint, and a flow that attributes cleanly to the site. The aggregated block sits below them with the same level of disclosure: what it is, why it is aggregated, and what is visible to the reader from it.

The contrast is the point. A reader who lands on the supply-chain page should be able to tell from a single read which categories of freight activity earn a per-site register at v1.0, which categories report in aggregate and why, and which categories are out of attribution at v1.0 with the path to inclusion stated rather than implied. The four aggregated categories on this page are scoped, not deferred. The point of the methodology is that scoping is itself a published act.

Related methodology

• Release cadence, editions, revisions, and the v1.0 exit
• Data window — what TRUST tells us, and what it cannot
• Operator league tables and Simpson’s paradox