Release cadence: editions, revisions, corrections, and how v1.0 exits

Gauge Intelligence publishes a dated public record, not a software product. Pages do not carry semantic version numbers. They carry an edition label, a last-updated date, and the date the next update is due. These are the conventions readers already recognise from the Office for National Statistics, the Office of Rail and Road, and the Bank of England.

The current edition is v1.0, the industry review edition. It is a published draft that invites correction from the operators, ports, terminals, regulators, and shippers it describes. The industry review window runs through March 2027. The next edition, v2.0, general availability, is declared when the four exit criteria set out below are met. This page is the reference for the schedule, the register conventions, and the exit test.

Editions, not versions

Every page on the archive carries the same three-line footer:

Edition · Last updated · Next update due

A quarterly release is identified by the period it closes: “quarter ending September 2026”, not “v1.1.0”. A monthly Network Rail period close is identified by the period: “P5 2026-27”, not “v1.0.5”. The arc from v1.0 (industry review) to v2.0 (general availability) is the only versioned scaffolding the archive carries, and it identifies the posture of the publication, not a point release.

The choice is deliberate. Dated editions tell a working reader two things a version number does not. First, when the figure was last touched. Second, when the figure will next be touched. A reader who needs to cite a number can read that the figure was last reviewed on 25 May 2026 and that the next scheduled review is the close of P5 (8 August 2026), and decide on that basis whether the citation is stable enough for their purpose. A version number requires the same reader to consult a changelog to reconstruct the same two dates.

Revisions and corrections: the ONS distinction

The Office for National Statistics separates two kinds of post-publication change. Gauge Intelligence adopts the distinction verbatim and applies it to every figure in the archive.

Revision. “Updates to previously published statistics, analysis or data that improve quality by incorporating improved methods, additional data sources or statistics that were unavailable at the point of initial publication.”

Correction. “An unplanned revision to any published ONS output because of the identification of an error.”

Source: ONS, National Accounts Revisions Policy (updated January 2026).

A revision improves a figure by adding evidence the original publication did not have, or by applying a better method the publication now uses. A correction repairs a figure that was wrong. The two kinds of change carry different disclosures and different reader obligations, and conflating them is the single most common failure of register in performance publishing.

A worked example. The first publication of a Felixstowe inbound A2F figure for P13 2025-26 used the partial-pool estimator with three months of trailing data. A subsequent edition, six months later, applies the same estimator to twelve months of trailing data and the figure moves by 0.6 percentage points. That is a revision. The estimator did not change; the evidence base grew. The original figure was correct for its evidence window; the revised figure is better. By contrast, if the first publication had attributed a delay to track that the Network Rail attribution review subsequently re-assigned to traction, the corrected figure repairs an error, and the page carries a dated correction notice, not a revision history entry.

Every page footer lists the page’s revision history and any corrections issued against it. The two lists are kept separate.

The four exit criteria for v2.0 general availability

The v1.0 industry review framing is falsifiable. It exits when all four of the following hold. Each is observable, dated, and audited against the production record. None is a date commitment. The point of stating them as criteria is that v2.0 cannot be declared while any one of them fails.

One: coverage threshold. At least 85% of train movements in the twelve-period rolling window have a populated lens-attributable entity-page home. Coverage here is the share of in-scope endpoints (the operator, the origin terminal, the destination terminal, and the corridor) for which a published register entry exists on the public archive. The denominator excludes two categories: sub-threshold endpoints with fewer than fifty movements per ninety-day window, and aggregated categories disclosed separately on the supply-chain page (Network Rail yards, nuclear logistics, defence logistics, and infrastructure-project flows). The carve-out is documented in full on the aggregate-coverage methodology page; the 85% figure is measured against the disclosed denominator.

Two: correction rate within tolerance. The rolling thirteen-period rate of corrections of error, in the strict ONS sense above, sits below a stated ceiling for three consecutive periods. The ceiling is fixed against the baseline observed at the close of P3 2026-27 (early July 2026). Until the ceiling is fixed, the correction rate is published every period without a tolerance judgement attached. A revision rate is published alongside the correction rate and is not subject to a ceiling, because revisions are how the record improves and suppressing them would degrade the archive.

Three: methodology stability. No edition in the preceding thirteen periods re-states historical figures as a consequence of a methodology change. Routine revisions that add evidence do not violate this criterion; only changes that move previously published numbers as a consequence of a new method do. Methodology consultations open at P6 (early October), close at P9 (early January), and any change with the power to move historical figures takes effect at P13. The consultation cadence mirrors the informal-engagement, formal- consultation, effective-date pattern established by S&P Global Commodity Insights for benchmark prices.

Four: named industry acceptance. At least three named industry signatories, across the freight operating company, port, and shipper segments, are willing to be quoted endorsing the dataset, or are actively citing Gauge Intelligence figures in their own commercial materials, regulatory submissions, or industry publications. Names and citations are listed on the published acceptance page; signatories may withdraw at any time. The point of requiring industry acceptance, not internal sign-off, is that the public record only earns the register it claims when the audience it describes is prepared to be associated with it.

The four criteria are stated as a conjunction. Coverage without acceptance is breadth without endorsement. Acceptance without coverage is endorsement of an incomplete picture. Methodology stability without a tolerable correction rate is rigour applied to errors. The exit is the joint condition.

The release calendar

Gauge Intelligence releases on three cadences, all anchored to the Network Rail period calendar (P1 begins on the first Sunday of April; P13 closes on 31 March).

A period close falls at the end of every Network Rail period (roughly every twenty-eight days). At each period close, the figures on every page that depends on the closing period are refreshed against the new period’s data. The refresh is rolling and unannounced. Where a figure has moved enough to alter the reading of a published claim, the page lists the change in its revision history.

A quarter close falls at the close of each Network Rail quarter (the periods that bound each Office of Rail and Road quarterly Freight rail usage and performance statistical release: end of June, end of September, end of December, end of March). At each quarter close, Gauge Intelligence publishes new entity register entries, new metric columns, new corridor coverage, and any substantive editorial extensions. The quarter close is the audited release boundary: the reader cites by quarter, the methodology stability test counts by quarter, and the corrections inbox processes within a quarter of receipt.

An annual close falls at the close of P13, roughly 31 March. The annual close carries the year’s methodology changes (consulted at P6, closed at P9, effective at P13) and the year’s published review. Where the methodology change re-states historical figures, the re-statement is a revision in the ONS sense, documented in the revision history on every affected page.

How corrections work

Anyone — an operator, a port, a regulator, a journalist, a shipper, a member of the public — can submit a correction by emailing the address on the licensing page. Each submission receives an acknowledgement within two working days and a substantive response within ten working days. Where the submission identifies an error, the page is corrected and the correction is logged in the corrections register with the submission date, the corrected figure, and the page footer last-updated date.

During the v1.0 industry review window the four launch-cohort pilots — GB Railfreight, Felixstowe, Doncaster iPort, and the Great Eastern Main Line — operate on an accelerated cadence: acknowledgement within one working day, substantive response within five. The accelerated cadence applies for the fourteen days immediately following first publication of a new edition of any of the four pilot reports.

Corrections to historical figures do not vanish. The corrections register is permanent and citable. A reader who relies on a figure that was subsequently corrected can read what the page said at the time of the original citation, why it was wrong, and what it now says.

Why this register

The choice to publish under a dated-edition register, with an industry review window, exit criteria, and an open corrections inbox, reflects a specific position on what a public record of rail freight performance should look like. It is closer to the Office for National Statistics, the Office of Rail and Road, the Bank of England staff working papers, and OpenSanctions than to any SaaS analytics platform. The pages do not invite trust on the strength of an assertion that the methodology is good; they earn trust by stating the methodology, publishing the corrections, and dating the next review. A working reader can audit any number against its evidentiary basis and can contest the figure if they have reason.

The industry review window is not a soft launch. It is the period during which the corrections that build the long-run register are received, processed, and documented. The exit to v2.0 is the moment at which the register has been sufficiently corrected by the industry it describes to warrant a general- availability claim. Until that moment, every figure carries the v1.0 edition label and an invitation to contest it.

Source

Office for National Statistics. National Accounts Revisions Policy. Updated January 2026.

Office of Rail and Road. Freight rail usage and performance, quarterly statistical release. Methodology and revisions policy sections.

S&P Global Commodity Insights. Benchmark methodology consultation process — informal market engagement, formal market consultation, effective date.