PlanOps method statement (RAMS) review standard

Version 2.1.0, 7 October 2026. First effective 30 July 2026 as v1.0.0. Published by PlanOps. Licensed under CC BY 4.0: use it, print it, adapt it, with attribution. The argument for the standard is in What a competent RAMS review actually tests.

What this is

A published, versioned standard for reviewing a subcontractor's risk assessment and method statement (RAMS) against the requirement the principal contractor issued for the activity. It defines twenty tests in four parts, the weight of each, the legal or project basis each one supports, and an arithmetic method that turns the reviewer's verdicts into a 0 to 100 score where every point lost is traceable to a named criterion.

It is written so that a health and safety manager can adopt it on paper without any software. It also describes the review framework used by PlanOps. This edition clarifies the published guidance; it does not announce a new machine-rubric release. An automated result must identify the rubric and evidence scope used for that run.

Version 2 makes the review proportionate to the risk of the work. A carpet fitter and a steel erector are still read against the same twenty tests, at the same weights, into the same bands. What changes is that a test the activity does not engage is recorded as does not apply, with the reason written down, rather than marked as a failure. Eight tests always apply, whatever the trade. The rules for the rest are under "Which tests apply" below, and the reasoning under "Why proportionate is the standard".

It is not a British Standard and it is not a legal opinion. Passing a test here is evidence towards managing a duty or requirement, not proof that the duty has been discharged.

The statutory references in this edition are mainly for construction work in Great Britain. References to Building Regulations Part 2A apply to building work in England. Wales has Part 2B duties and competence requirements effective from 1 July 2026; check the applicable regime and transitional arrangements. Building-regulations competence concerns compliance of building work and is distinct from construction workplace safety under CDM. Different legislation applies in Northern Ireland.

How to use it on paper

The working checklist fits on two sides of A4: download the two-page checklist (PDF), print it, and put it next to the submission.

1. Have the requirement you issued for the activity to hand: the risk category, the legislation you cited, any permit, any temporary works reference, the activity description, the risk level, and your organisation's own rules if any.
2. Write down the hazard profile of the work: each hazard category it involves and how serious it is. Take it from the requirement where you recorded one. Where you did not, rate it yourself from the work described, and never from the submission's own rating (see "Which tests apply" below).
3. Decide which tests apply before you mark any of them. Eight always do. For each of the others, ask whether this activity engages it. Where it does not, record does not apply and one sentence saying why. Where it does, the test applies, and a RAMS that is silent on it has not met it.
4. For every test that applies, record a verdict: met, partially met or not met. Every criterion must carry either a verdict or a reasoned does not apply.
5. For each criterion that is partially met or not met, record a severity, judged as the consequence if this work, with these hazards, proceeds as written: critical, major or minor. Critical means work under this RAMS could kill or seriously injure someone (no rescue plan, no temporary works design). Major means a material gap that must be closed before work starts. Minor means a limited deficiency that would not, on its own, materially affect safe execution.
6. Score it using the method below, or simply apply the caps: any critical not met, serious gaps, send it back; any critical partially met, significant gaps; any major finding, at best gaps to close.
7. Return the verdicts to the subcontractor with the review, including the reason for every test you marked as does not apply. The list of what was tested, and of what was not and why, is the review.

The four parts and twenty criteria

Weights sum to 100. The "evidence basis" column is a guide to why the criterion matters. It may name a direct legal duty, recognised guidance or good practice, or a project or contract requirement. It should not be read as saying the law requires that exact RAMS heading or wording. Abbreviations: MHSWR is the Management of Health and Safety at Work Regulations 1999; CDM is the Construction (Design and Management) Regulations 2015; HSWA is the Health and Safety at Work etc. Act 1974; Part 2A is Part 2A of the Building Regulations 2010 (England, competence, from 1 October 2023).

Part 1: Is the risk assessment real? (18 points)

Part 2: Does the method statement say enough? (56 points)

Each criterion asks the same shape of question: is this section present, specific and adequate?

Part 3: Does it answer the requirement you issued? (22 points)

Each criterion compares the submission against the requirement for this activity.

Part 4: Does it meet your own rules? (4 points)

Two conduct rules apply to Part 4 and are not scored: cite each organisation requirement by its exact reference when it is used as a test; and where no organisation requirements were issued, do not invent any, mark criterion 20 as does not apply.

Practical checks within the criteria

These examples explain the existing hazard, control and emergency criteria. Apply them where the activity presents the hazard; they are not extra scored criteria.

• People and actions: identify who may be harmed and how, existing controls, further actions, who will implement them and by when. HSE's assessment guidance sets out these questions. Record action owners and deadlines for outstanding actions; do not invent a statutory requirement for a particular table layout.
• Residual risk: where a numerical rating is used, check that the stated reduction is supported by specific, workable controls. PPE and competence may contribute, but mentioning either does not by itself justify a low score. Assess the suitability and effectiveness of the measures for the hazard.
• Work at height: follow avoidance, fall prevention and mitigation, considering collective protection before personal protection. Where fall arrest is used, check suitable rescue arrangements.
• Electrical work: check dead working and safe isolation, including preventing re-energisation and proving dead. A live-work proposal needs a specific justification against the applicable legal conditions and precautions; a competent electrician alone is not the justification.
• Confined spaces: consider avoiding entry first. If entry is necessary, check the safe system and effective rescue arrangements. Reliance on emergency services alone is insufficient.
• Hazardous substances: check task-specific exposure assessment and controls. A safety data sheet is an input to a COSHH assessment, not its replacement. Apply the separate asbestos and lead requirements where those hazards are present.

Supporting evidence and the reviewer decision

A reference to another document is not evidence that its contents were assessed. Record the review's evidence scope separately from whether a criterion applies.

Where an applicable control is not evidenced, report that gap with a severity based on the hazard. Do not credit an unseen rescue plan or design merely because the RAMS names it. If a reading failure prevents a criterion being assessed, apply the unavailable-score rule; do not disguise a technical failure as a safety finding. If an ancillary file is not needed to reach a finding, explain that limited scope rather than assuming every missing reference invalidates the whole review.

Before accepting, the reviewer should establish how material findings are resolved: revised RAMS, additional evidence, or a reasoned disagreement tied to the activity and site. A criterion can remain applicable even where further evidence resolves its finding. Record the reviewer, date, RAMS revision, reasons and evidence relied on; preserve the original AI assessment alongside that decision and any later assessment. Keep this evidence with the review so that another competent person can understand the basis of the decision.

Acceptance does not replace required permits, briefings, checks or continuing supervision, and does not transfer statutory duties. Obtain or verify essential evidence before the affected activity starts. The contractor and principal contractor retain their respective responsibilities for implementing and monitoring safe work.

Which tests apply

A criterion that does not apply leaves the denominator and the remaining weight is renormalised, so a RAMS is neither rewarded nor punished for a test its work does not engage. That is the proportionality lever, and because it can also be a route for a real gap to disappear, version 2 puts four rules around it.

1. Eight tests always apply. Criteria 1, 2, 3, 5, 6, 7, 11 and 13 apply to every RAMS, whatever the trade: a real risk assessment, the foreseeable hazards, a stated scope, competent people and suitable equipment, task-specific hazard identification, workable controls, training and information, and monitoring. Criterion 18 also always applies whenever a requirement was issued that records an activity description; a requirement silent on the activity records no scope to conform to, so 18 does not apply with that reason. Marking any of these as does not apply is not a judgement the reviewer can make. It leaves the criterion unassessed, and the coverage rule then withholds the score. The eight weigh 50 points between them, so a scored review never rests on less than half the standard.

2. The rest apply when the activity engages them.

The distinction that matters is between an activity that does not engage a test and a RAMS that does not evidence it. For a floor-covering installation confirmed to involve no temporary works or lifting, criterion 9 does not apply. For steel erection requiring temporary bracing, an absence of evidence for stability arrangements can be a critical failure. Judge the work and the evidence, not the trade name. Silence in the document is never a reason for does not apply.

Some of these the record decides. Criterion 16 considers the permit requirement and supplied references; criterion 17 considers temporary works referenced or linked to the requirement; criterion 20 considers issued organisation rules. Apply the record-based rules below without inventing missing project requirements. The others are judgements against the hazard profile, which is why the profile is written down first.

3. Every does not apply carries its reason. One sentence, recorded with the verdict and returned with the review, saying what about this activity takes the test out of scope: "no temporary works are involved in a floor-covering installation", not "not relevant". A does not apply with no reason counts as unassessed, and the coverage rule withholds the score rather than letting an unexplained exclusion quietly shrink the denominator.

4. Severity follows the hazard profile. The severity definitions have always been written as the consequence if the work proceeds as written. Version 2 says what that means in practice. The same finding carries a different severity on different work: missing effective rescue arrangements can be critical for confined-space entry or work relying on fall arrest. A minor documentation gap is only minor where evidenced arrangements are adequate for the actual hazards; being a ground-floor fit-out does not establish that. Consider fire, substances, access and the people affected. The factors, the caps and the band meanings do not change with the trade. What changes is which gaps are material.

Where the profile comes from. Take it from the requirement you issued. Where no requirement recorded one, a reviewer may rate the hazards from the work the submission describes, on the inherent hazard of that work and never on the document's own rating, or the RAMS under review would set the benchmark for its own controls. A profile taken from the submission cannot then be used to test it: criterion 14 does not apply, because conformity to a profile drawn from the document proves nothing, and hazard coverage is judged under criterion 2, which always applies. Where no profile is recorded and none can be rated, criteria 14 and 19 do not apply, and the review says so in those words rather than assuming a medium risk nobody assessed.

Why proportionate is the standard

Proportionality is not a softening of the review. It is what the law asks for.

Regulation 3 of MHSWR requires a suitable and sufficient assessment of the risks. What is sufficient for a carpet fit and what is sufficient for steel erection are different things, because the risks are different; an assessment that is suitable and sufficient is proportionate to the risk by construction. HSE's risk assessment guidance puts the same point plainly: you are not expected to eliminate all risks, but to do everything reasonably practicable, balancing the level of risk against the measures needed to control it.

There is no general statutory requirement for a document titled "RAMS". Risk assessment and safe planning duties still apply, and some activities require written arrangements. HSE's construction guidance identifies demolition, dismantling and structural alteration as requiring those arrangements before work begins. For many other activities, a method statement is a practical means of planning, managing and communicating safe work. A review should distinguish legal duties, recognised guidance and project requirements, and apply each to the work described.

The version 1 failure showed the cost. An adequate carpet-fitting RAMS was marked not met at major severity on temporary works, permits and referenced temporary works, because nothing told the reviewer those tests could fall away. The major-finding cap then placed it in gaps to close, next to documents with real material gaps, telling the site that work should not start until temporary works on a carpet fit were closed. Worked example 4 below runs the numbers.

The weights, factors, caps and band thresholds are shared across trades. Proportionality comes from which tests apply and how severe each gap is for the work. Equal scores do not establish equal site risk or equal review coverage: compare the applicable criteria, evidence scope, findings and rubric version as well as the number.

When software runs the review

HSE set out its regulatory approach to AI on 12 June 2026. Existing health and safety law is goal-setting, so it applies to the use of AI at work without new legislation, and the principle it applies is the familiar one: those who create risks are best placed to manage and control them "in a sensible, proportionate, and pragmatic way". Among its planned work for 2026/27, HSE lists research into cognitive offloading in AI-augmented safety decision making. (HSE's regulatory approach to artificial intelligence)

For a RAMS review that means using software changes nothing about who holds the duty. The principal contractor's review is still the principal contractor's, and the reviewer still needs to be able to read, check and overturn every verdict. This standard is built for that. The score is arithmetic, never asked of a model. Where the record settles whether a test applies, the workflow decides it and the model cannot override it. Every does not apply carries its reason on the face of the review, so the one lever that can make a gap disappear is also the one most visible to the person signing it off. A review whose exclusions cannot be read is a review that invites someone to stop reading.

The document-only subset

The twenty criteria divide by what they need in front of the reviewer, and any review claiming this standard must say which set it ran.

Criteria 1 to 13 are document-only. They assess the RAMS and any supporting evidence actually available to the review: whether the risk assessment is real and whether the method statement says enough. Together they carry 74 of the 100 weight.

Criteria 14 to 20 are comparison criteria. Every one tests the submission against something the reviewer must hold: the requirement issued for the activity (its risk category, cited legislation, permit, temporary works references, activity description and risk level) for 14 to 19, and the organisation's own rules for 20. When no requirement is held, criteria 14 to 19 do not apply; criterion 20 also does not apply when no organisation rules were issued. When a requirement exists but has blank fields, use the per-criterion rules below. Those rules allow supplied permit references for criterion 16 and an independently judged inherent hazard profile for criterion 19. They do not allow the reviewer to invent a project requirement or accept the RAMS's own residual rating as its benchmark.

What each comparison criterion compares against. So that the gates are readable on paper, and in the same words the workflow uses:

Where software runs the review, the workflow settles these rather than the model: "criteria 14 to 19 are not applicable when no RAMS requirement is held, with a separate basis naming each missing comparison; criterion 20 is also not applicable on a document-only run when no organisation requirements were issued. When a requirement is held, the 2.0.1 gates for criteria 14 to 18 still follow its recorded facts. An expected requirement that cannot be read stops the audit before model judgement. The model cannot override a workflow-established not-applicable verdict." A requirement that exists with blank fields is not the same thing as no requirement at all: the fields it does record are still compared.

A review run without the requirement is therefore a document-only review: at most thirteen criteria over an applicable weight of at most 74, never fewer than the eight that always apply, which weigh 50, and stated as such. It is a legitimate first pass and an incomplete review, and the result must say which it was. This is exactly how the free review at ramsreview.co.uk runs: it reads one document with no sight of any project, so it assesses the thirteen document-only criteria and reports criteria 14 to 20 as not applicable. Inside a project, all twenty criteria are considered, with applicability determined from the work and available project requirements.

Any result from a document-only review should carry the subset on its face, not in a footnote, ahead of the score: "Document-only review: 13 of 20 tests assessed, applicable weight 74" where every document-only test applied, and the actual count and weight where the activity took some of criteria 4, 8, 9, 10 and 12 out of scope. A document-only 92 and a full-review 92 are not the same level of scrutiny, and the result is the place to say so.

A result may state that coverage in plainer words on screen, provided the statement above appears on the record that goes with it. The wording is fixed for the record; the screen may say the same thing in the language of the site it is read on.

Scoring method

The score is arithmetic over the verdicts. The reviewer is never asked for a number.

Verdict factors.

Severity factors.

Formula.

applicable_weight = sum of weights of criteria that apply
deduction(i)      = weight(i) x verdict_factor(i) x severity_factor(i)
score_before_caps = 100 x (applicable_weight - sum of deductions) / applicable_weight
score             = round( min(score_before_caps, cap) )

where cap = 100 when no cap condition below applies

Because no deduction can exceed its own weight, the score is inside 0 to 100 by construction.

Caps. Applied after the arithmetic, lowest cap wins.

Coverage rule. Every criterion must carry a verdict. A score derived from a partial assessment is not a score: when any criterion is unassessed, report the score as unavailable and name the criteria that were missed. A does not apply with no reason, or on one of the tests that always apply, counts as unassessed.

Floor. Because the eight tests that always apply weigh 50, applicable weight is never below 50 on a scored review.

Unavailable rule. When the review could not be read or completed, the score is null with a stated reason. Zero is never used to mean "could not assess". Zero is a real score, reserved for a submission that fails every applicable criterion at critical severity.

Bands

The score is a review indicator, not a percentage of safety or legal compliance. A score of 49 may result from a critical-failure cap rather than a deduction to 49. Acceptance must address the findings and evidence, not just the band.

Band names describe the RAMS overall; they are not descriptions of the severity of individual findings. A single major finding places the whole document in the gaps-to-close band, because the top band is defined by the absence of material gaps.

Worked examples

1. Everything met, no organisation rules issued. Criteria 1 to 19 met; criterion 20 does not apply. Applicable weight 96, deductions 0. Score 100. No material gaps.

2. One major gap only. Everything met except emergency arrangements (criterion 8, weight 7), not met at major severity. No organisation requirements were issued, so criterion 20 does not apply. Deduction 7 x 1.0 x 0.6 = 4.2 of 96 applicable, which is 95.6 before caps. The major-finding cap brings it to 89. Gaps to close. A major gap cannot sit in the top band.

3. The steel-frame erection RAMS. A synthetic example used to design and test the method (the subcontractor and site are fictional). Twelve-storey steel frame, requirement titled "Tower crane operations, steel frame erection", citing LOLER 1998, CDM 2015 and BS 7121-1 and BS 7121-5, requiring a lifting permit and referencing the crane foundation temporary works, adjacent to an occupied office building with a noise restriction.

Applicable weight 96. Deductions 47.2. Score before caps 50.9. Five criteria not met at critical severity (6, 8, 9, 14, 15), so the cap applies: 49. Serious gaps. Every variable criterion is engaged by steel erection next to an occupied building, so version 2 changes nothing here. Proportionality does not rescue a document with real gaps.

4. A carpet fit, under version 1 and version 2. Also synthetic. Floor coverings on the second floor of an occupied office, adhesives in use, no temporary works, no permit, no organisation rules. The reviewer has checked applicable site first-aid, alarm and evacuation arrangements, and no specialist rescue is needed. The RAMS has a minor cross-reference omission in its emergency section: partially met, at minor severity on that evidence. Without adequate evidenced arrangements, that severity would need reassessing.

Under version 1, nothing allowed temporary works or permits to fall away, so criteria 9, 16 and 17 were marked not met at major severity. Deductions 3.6 + 1.8 + 1.8 + 0.875 = 8.1 of 96 applicable, which is 91.6 before caps. Three major findings, so the major-finding cap applies: 89. Gaps to close, with temporary works among the gaps.

Under version 2, criteria 9, 16, 17 and 20 do not apply, each with its reason: no temporary works or lifting in a floor-covering installation; no permit required, named or referenced; no temporary works referenced or linked; no organisation requirements issued. Applicable weight 84. Deduction 7 x 0.5 x 0.25 = 0.875, which is 99.0 before caps, and no cap applies: 99. No material gaps. Give the same carpet fitter a genuine major gap, such as no COSHH assessment for the adhesives, and the major-finding cap still brings it to 89, as it should.

The two scope criteria

Criteria 3 and 18 both concern scope and deliberately test different things. Criterion 3 sits in Part 2 and asks what every criterion in that part asks: is the section present, specific and adequate. Criterion 18 sits in Part 3 and does what every criterion in that part does: compares the submission with the requirement. The failures are different and so are the remedies. A vague scope means rewrite the method statement; a mismatched scope means this is the wrong RAMS for the activity. Up to v1.0.0 the two were worded so similarly that a scope mismatch was double-counted in practice; v1.1.0 corrected the wording rather than merging the criteria, so no ids, weights or scores changed.

Changelog

The published edition and machine rubric have separate versions. Edition 2.1.0 updates the guidance and checklist, including criterion 18's activity-description condition already recorded in rubric 2.0.1. It does not change criterion ids, weights, factors, caps or band thresholds, or release a new machine rubric. The previously published v2.0.0 checklist remains available. Each automated run must retain its own rubric version; publication of this edition does not retrospectively change an assessment.

Criticism and adoption

If a weight, a test or a duty reference is wrong, tell us. Changes that alter the meaning of a stored score (renumbering, reweighting, new criteria) are published as a new major version with the old one kept readable. Clarifications are minor versions. Contributors are credited in the changelog.