Peer review, sign-off and independent audit of government analysis are all optional under the government's own standard for it, only naming someone responsible is compulsory. The current edition also cannot describe itself correctly, reusing another standard's authorship and changelog word for word.
Estimated reading time: 8 minutes
In short. GovS 010: Analysis is the government functional standard that sets expectations for "all government analysis carried out by anyone within or on behalf of government", in every department and arm's length body, whether published externally or used inside a decision no one outside government ever sees [1] [2]. Its principles state that analysts "shall ensure... ongoing quality assurance and continuous improvement is embedded in the way analysis is carried out" [2]. Every mechanism the standard actually describes for delivering that, peer review, independent audit, verification and validation, formal sign-off, is written as "should", not "shall" [2]. The current edition's own front matter, version 2.2, issued 26 August 2025 and still live and unamended today, describes a different standard's rebrand and credits authorship to "the People Function", when the document is Analysis, published by the Government Analysis Function [2] [1]. Three documented mistakes: four stars.
GovS 010 sits inside a suite of government functional standards, cross-referenced against GovS 002 (Project delivery), GovS 005 (Digital), GovS 007 (Security), GovS 008 (Commercial) and GovS 011 (Communication), and it names its own audience precisely: "permanent secretaries, directors general and chief executive officers of arm's length bodies and suppliers", plus "users and producers of government analysis, including non-analysts and external consultants" [2]. Its scope clause leaves nothing out: it applies "regardless of analytical methodology or technique used", to analysis "informing decisions on policies, project delivery and operational services and informing the public" [2]. This is the document that is supposed to stand behind every number a minister quotes and every model a department builds.
A shall with no shall behind it
Section 2, Principles, opens with a direct instruction: "Those engaged in analysis shall ensure... ongoing quality assurance and continuous improvement is embedded in the way analysis is carried out" (p.5) [2]. The standard defines its own vocabulary on the inside cover: "shall" is mandatory, "should" is advisory, a distinction it says exists precisely so mandatory and optional elements are never confused [2].
Section 4.2, Assurance of analysis, is where that mandatory principle is supposed to become a working mechanism. It sets out three levels of assurance, "by, or on behalf of operational management", "by, or on behalf of senior management, independent of operational management: analytical peer review", and "by independent bodies: analytical audits" (p.8) [2]. Every one of those three levels is introduced with "Organisations should have a defined and established approach to assurance... typically... on at least three separate and defined levels" (p.8) [2]. Should, not shall.
The pattern holds through the rest of the document. The analytical assurer role exists specifically to confirm "appropriate and sufficient analytical quality assurance has taken place" (p.9), but the design "should" be checked by the assurer, not shall [2]. Formal sign-off is the same: "The analysis commissioner should formally approve the results of the analysis for dissemination" (p.14), and they "should be satisfied that the analysis has been undertaken competently and the results are valid" (p.14) [2]. Analytical verification and validation, the section whose whole purpose is stated as assuring "that the analysis has been conducted properly", is introduced with "should be proportionate to the purpose and constraints of the analysis" [2].
Search the document for what is actually mandatory, and two things qualify: "Each organisation shall have a senior officer accountable for its analysis activities" (p.8), and "Roles and accountabilities shall be defined... and assigned to people with appropriate seniority, skills and experience" (p.9) [2]. Naming a person is mandatory. Nothing that person actually does, peer review, audit, sign-off, verification, is. The document contains no consequence, non-compliance or escalation clause of any kind for an organisation that skips every "should" while keeping the one named officer the standard requires. The Principles page's "shall" and the Assurance page's "should" describe the same activity, in the same document, with nothing in between them.
A citation seven months out of date
GovS 010 points analysts to HM Treasury's guidance for the substance of quality assurance: "The Aqua Book provides detailed guidance on producing quality analysis for government" and "The Green Book... set[s] requirements and provide[s] guidance on appraisal and evaluation", both again on a "should be followed" basis rather than shall (p.16) [2]. Annex A's own reference list cites it precisely: "4 HM Treasury, The Green Book: Central government guidance on appraisal and evaluation (2022)" (p.20) [2].
That edition no longer exists as live guidance. HM Treasury replaced it on 5 February 2026 with a new edition it describes as "radically shorter and simpler", checked live today with no reference anywhere back to GovS 010 [3]. GovS 010 version 2.2 was itself issued after that date, 26 August 2025 is before it, so the 2022 citation was already stale the day version 2.2 went live, and it has not been corrected in the seven months since the rewrite. This site's own teardown of the 2026 Green Book, published 13 September 2026, found that its central safeguard against cost blowouts, a mandatory, explicit adjustment for optimism bias, was applied correctly to High Speed Two's Phase 1 approval and still failed to catch a budget swing that later led the same department to rate its own 2020 approval as poor value for money. GovS 010 defers analytical quality assurance to that document on an advisory basis, while citing an edition of it that had already been withdrawn.
The claims, tested
| The document's own words | What checking the citation shows | Verdict |
|---|---|---|
| "Those engaged in analysis shall ensure... ongoing quality assurance and continuous improvement is embedded" (p.5) [2] | Every named assurance mechanism, peer review, independent audit, sign-off, verification and validation, is written as "should" throughout sections 4.2, 5.3, 5.6 and 6.6 [2] | A mandatory principle with no mandatory mechanism behind it |
| The standard's own front matter states its edition history: "Version 2.1 replace version 2 GovS 003, now titled 'People' instead of 'Human Resources'... Produced by the People Function" (inside cover) [2] | GovS 003: People is a separate, live standard, last updated 30 June 2025, whose own title change this paragraph describes [4]. GOV.UK's own metadata for GovS 010 names the publishing organisation as the Government Analysis Function and records the real version 2.2 changes as "better linking with other functional standards, a new collaborative working clause and other minor improvements" [1] | The current edition of an analysis standard misdescribes its own authorship and its own changelog |
| "The Green Book... set[s] requirements and provide[s] guidance on appraisal and evaluation" and should be followed, citing "The Green Book... (2022)" in Annex A (p.16, p.20) [2] | HM Treasury replaced the 2022 Green Book on 5 February 2026 [3], six months before GovS 010's own current edition, version 2.2, was issued on 26 August 2025 in the case of the citation predating the rewrite, and unrevised ever since | An advisory citation, to an edition that no longer exists |
The mistakes, counted
The standard's one mandatory quality principle has no mandatory mechanism (1). Section 2 states, using "shall", that ongoing quality assurance must be embedded in every piece of government analysis. Every mechanism the rest of the document describes for actually doing that, peer review, independent audit, formal sign-off, verification and validation, is written as "should". The only things GovS 010 actually mandates are naming a senior officer and defining roles, neither of which requires any check to happen at all.
The current edition misattributes its own authorship and changelog (2). Version 2.2's front matter, live since 26 August 2025 and unamended for over a year, states that "Version 2.1 replace[d] version 2 GovS 003, now titled 'People' instead of 'Human Resources'" and that the document was "Produced by the People Function". GovS 003: People is a separate standard with its own edition history; GOV.UK's own page for GovS 010 names the Government Analysis Function as publisher and gives a different, accurate account of what version 2.2 actually changed.
Its citation to the Green Book is stale (3). GovS 010 defers the substance of appraisal quality assurance to the Green Book on a should-be-followed basis, and its own Annex A cites the 2022 edition by name. HM Treasury replaced that edition on 5 February 2026. The current edition of GovS 010 has not been updated to reflect it.
Credit where due
The substance of GovS 010, read past the shall-should split, is a careful document. Its account of the analytical cycle, from initial scoping through to sign-off, tracks how good analysis is actually produced, and its roles, the analysis commissioner, the analytical assurer, the analyst, are clearly defined with real accountabilities attached, even where the standard stops short of mandating that those accountabilities be exercised in every case. The advisory framing is also defensible in principle: government analysis ranges from a one-page ministerial briefing to a multi-year appraisal, and a single mandatory assurance process applied identically to both would be a bad standard in its own right. None of that explains why the current, live edition of the government's standard for analytical rigour carries someone else's changelog on its own inside cover, uncorrected for over a year.
Verdict
Four stars, from three documented mistakes. GovS 010 sets one binding rule, that quality assurance shall be embedded in government analysis, and then hands every mechanism capable of delivering it to a "should". The one part of the document that should have been effortless to get right, describing itself, names the wrong function and the wrong changelog. A standard that exists to make government analysis checkable has gone over a year without anyone checking its own front page.
Sources
- Cabinet Office and Government Analysis Function, "Government Analysis Functional Standard" (publication landing page, version 2.2, last updated 26 August 2025, checked 25 September 2026). https://www.gov.uk/government/publications/government-analysis-functional-standard--2
- Government Analysis Function, "GovS 010: Analysis, Version 2.2" (PDF, dated 26 August 2025, checked 25 September 2026). https://assets.publishing.service.gov.uk/media/68ad6757969253904d1557f7/GovS_010_v2.2.pdf
- HM Treasury, "The Green Book: appraisal and evaluation in central government" (publication landing page, last updated 5 February 2026, checked 25 September 2026). https://www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-government
- Cabinet Office and People Function, "Government Functional Standard GovS 003: People" (publication landing page, last updated 30 June 2025, checked 25 September 2026). https://www.gov.uk/government/publications/government-functional-standard-govs-003-human-resources