
By role · Internal auditor
Your standards tell you how to audit. They do not tell you what to look at.
The auditing standards govern independence, evidence and reporting. Establishing the criteria is your job. The procurement guidance cited here stops at contract closure; criteria past that point come from the relevant asset, accounting and review requirements.
Who this is for
You have been asked to give assurance on a capital programme, or on an entity that is building assets it will shortly have to run. You are not the engineer, and none of your standards makes you one.
What you already work under
Standards governing your work
Government Auditing Standards, 2024 revision, GAO-24-106786, effective for performance audits beginning on or after 15 December 2025
Nine chapters: the foundation and use of the standards, general requirements for compliance, ethics and independence, competence and continuing education, quality management and peer review, and then the conduct and reporting of financial audits, attestation engagements and performance audits. Not one of them names a subject. GAGAS defines planning to include determining objectives, scope and methodology and establishing criteria to evaluate matters subject to audit, which puts the criteria on the auditor rather than in the standard.
Global Internal Audit Standards, Institute of Internal Auditors, effective 9 January 2025
Five domains and 15 principles covering the purpose of internal auditing, ethics and professionalism, governance of the function, its management, and the performance of services. The same shape as GAGAS: how the function is run and how an engagement is conducted, not what a capital programme should be measured against.
INTOSAI GUID 5280, Guidance for Audits of Public Procurement, 2022, free in full
The closest thing to a published, global question set for this ground. Three phases, pre-tendering, tendering and post-award, with annexes carrying roughly a hundred possible audit questions and a table of integrity risks at each step. It is also explicit about its own limits: paragraph 62 says the questions are a menu to choose from, are not exhaustive, and need to be completed with the applicable or designed audit criteria.
Read the sources
- Government Auditing Standards, 2024 revision, GAO-24-106786, Comptroller General of the United States
2024 revision; read 2026-09-04. Free in full. Effective for performance audits beginning on or after 15 December 2025. - Global Internal Audit Standards, Institute of Internal Auditors
2024 Global Internal Audit Standards; read 2026-08-20. - INTOSAI GUID 5280, Guidance for Audits of Public Procurement, 2022
2022; read 2026-09-04. Free in full. Ten annexes, including possible audit questions for each of the three procurement phases.
Where that stops
GUID 5280 ends where the contract does. Its last post-award question asks whether the contract was closed only once the contracting authority formally recognised that all deliveries were complete and payments were released. Everything this publication is about starts one step after that: the asset exists, somebody has to run it, it has to appear in a register at a value somebody can defend, and it creates operating, maintenance, renewal and disposal costs. The only place the asset register appears in GUID 5280 at all is in the integrity risk table, as something that might be stolen from before it is recorded. GUID 5280 does not provide criteria past that line; they come from the relevant asset, accounting and review requirements. What follows is a set you can argue with.
The reading path
The material, on the same lifecycle spine as everything else.
Whether the decision to build was made on evidence that would survive being read back, and whether the governance that approved it could have said no.
What to examine
- Need and outcomeIs there a stated outcome this asset is meant to produce, written down separately from a description of the asset itself?
- The decision-making frameworkAre the criteria by which options were chosen written down, and is it recorded who was entitled to choose?
- Definition maturity and the estimateDoes the estimate state the class it was produced at, and does that class match the scope definition actually achieved?
- Sanction and the authority to stopWho is entitled to stop this, and have they ever been given information that would let them?
Instruments
- The question bank
Every question this publication asks, drawn from the framework and the articles, filtered by stage and by source, printable.
- Lifecycle cost sensitivity
Take three real costed buildings and watch the construction share of lifecycle cost move from 66% to 7% while the data stays exactly the same.
Related writing
- A gate that has never been failed is not a controlAssurance before anything is built costs less and changes more than assurance anywhere else in the lifecycle. It is also the assurance that is hardest to buy, and the reason is not that owners are careless. It is that inception produces very little of the sort of evidence an assurance function knows how to examine.7 min read
- Weighting and scoring is the method the guidance warns againstThe standard option appraisal is a table. Options down the side, criteria across the top, weights, scores, a total, a winner. The Green Book names that technique and recommends against it, on the grounds that it lacks an objective basis and reduces transparency. What it recommends instead does not rank options at all. It builds them.7 min read
- The largest contributor to optimism bias is the document that starts the projectHM Treasury publishes uplifts by project type, from 24% on a standard building to 200% on equipment and software. The instruction is to start at the upper bound and come down only against verified evidence. The single largest contributory factor it names is not the contractor, the ground or the weather.7 min read
- An estimate is a measure of how well the scope is definedEarly cost estimates are produced as ranges and quoted as commitments. The classification systems behind them say something narrower and more useful. Accuracy is governed by the maturity of the scope definition, and the published ranges exclude the project types a large mixed-use development is made of.8 min read
- Whether it was worth doing is the least evaluated questionThe UK government reviewed evaluation across the 227 largest projects it runs, together worth £834 billion. A third have a plan capable of finding out whether they worked. Of the three kinds of evaluation, the one asking whether the benefits justified the cost is the weakest, and the reason is that it has to be designed before the project starts.7 min read
Whether the reported position is the real one, and whether the assurance already being performed adds up to coverage or only to activity.
What to examine
- Baseline integrityIs the baseline currently being reported against the one that was sanctioned, and if not, is every step between the two documented?
- Change, planned and unplannedAre changes assessed before they are made, and does anyone examine changes that were made without being assessed?
- Forecast credibilityDoes the forecast reflect what is now known, or does it reflect what was committed?
- Externally provided processes and servicesWhere activities that affect the asset have been placed outside the organisation, has the organisation retained the ability to know whether they were done?
- Competence of whoever is assuringDo the people performing this assurance understand the thing they are examining, and where they do not, is the gap covered by someone who does?
Instruments
- Earned value calculator
From planned value, earned value and actual cost, the variances and indices; add the budget at completion for the estimates at completion. Formulas shown and cited.
- Reporting history check
It does not ask whether the project is healthy. It asks whether the reporting on it has ever produced a signal, across the whole history rather than this month.
Related writing
- Every programme performs well against a baseline it can moveA government builder completed 36 major construction projects and routinely met its cost and schedule goals. It had rebaselined 25 of them, and its reports did not say so. When asked why they had moved a baseline, 4% of project managers gave their own cost or schedule performance as the reason.9 min read
- A forecast that never moves is not stable, it is unexaminedDelivery assurance measures progress against a baseline. Testing the baseline is a different exercise, and it is the one the monthly report does not contain. A forecast that holds flat and then steps was not stable, it was unexamined, and two numbers usually available in an afternoon will show which.6 min read
- Programme risk is not the sum of its functionsA capital programme does not simply contain larger versions of procurement, payment and resourcing. It creates chains that run through all of them, and a chain belongs to no function. That makes under-coverage a design problem in the audit plan rather than a question of materiality, and it has a method attached.11 min read
- Engaging an engineer satisfies the standard. Deferring to one does not.The 2024 Standards require an internal audit function to obtain the competencies it does not have, which reads as permission to bring in an engineer. That is the easy half. The obligations that decide whether the exercise was worth commissioning all start after the engineer has arrived.7 min read
- You can place the activity outside. You cannot place the accountability thereClause 8.3 was retitled in the 2024 edition from Outsourcing to externally provided processes, products, technologies and services. The committee says the requirements did not substantially change. The four nouns still describe a category of arrangement wider than what outsourcing usually names, and the wider category is the one nothing in the audit plan is pointed at.7 min read
- Your contingency and their contingency are different moneyTwo bodies an owner is likely to cite define contingency and management reserve in opposite directions. One says contingency is for unknown unknowns and sits with the owner. The other treats it as the allowance for what experience says will happen. Both are defensible, and an owner and a contractor using the word across a table are frequently discussing different money, held by different people, against different risks.7 min read
- Reliability can be specified like mass. Building owners do not.The committee that writes the world's dependability standards says reliability should be specified the way dimensions and mass are. Rail has had a published process for specifying and demonstrating it since 2002. The buildings sector's reference specification for what a contractor hands over asks for the manufacturer's projection of how many hours a repair will take, and never how often one will be needed.10 min read
- The score that is not a numberA technical score on a qualitative criterion records what evaluators judged. The weighting decides how much of the award that judgement controls; it does not turn the judgement into a measurement. What makes the score answerable is a record that a reader outside the room can test against the solicitation and the proposal, and the instruments examined here require different parts of that record.9 min read
- An extension of time is tested against records made during the delayA claim for more time turns on three separate questions. Was the contract's notice procedure satisfied? Is the event one for which the contract allows an extension, or whose risk it otherwise places on the employer? Did it actually delay completion? Money is a separate claim, with its own proof of causation and amount. Each question is later tested against what was written down while the event was happening, by whichever party has to prove the point. Records do not create entitlement, a record of an event does not prove that it delayed completion, and the prevention principle does not rescue a claim under every contract or every governing law.14 min read
Whether the conditions for taking over were met or waived, and whether anybody tested the receiving organisation rather than only the plant.
What to examine
- Information transfer against stated requirementsWas what was delivered tested against what was asked for, by someone who will have to use it?
- Knowledge, as distinct from informationHas anyone captured why the asset is the way it is, and what was accepted rather than resolved?
- Taking over, and what was waived to achieve itWere the contractual conditions for taking over actually met, and where they were not, who agreed to proceed and on what record?
- Operational readiness of the organisationDoes an organisation exist that is capable of operating this asset on the day it is handed over, with people in post?
Instruments
- Operational readiness check
Put the questions a readiness review will ask, condition by condition, and name the blockers with their references. The percentage counts your answers; the blockers apply a rule.
- Turnover package composer
What has to change hands at taking over, grouped by what it is for rather than by who produced it, with what depends on each item and how to tell whether it is any good.
- What follows the certificate, and when
Enter one completion or taking-over date and get every consequential date under your contract regime, each computed from the clause that sets it.
Related writing
- At handover, every review arrives after the leverage has goneAssurance at this stage is triggered by the event that ends the ability to do anything about it. A review three months before taking over can still change what gets produced. The same review afterwards can only describe what is missing, and describing what is missing is not assurance, it is an inventory of the loss.6 min read
- Completion describes the asset. Readiness describes the owner.Commissioning is the one handover instrument that tests rather than documents, and it tests the building against the owner's stated requirements. Everything on the other side of the boundary, whether the owner can actually run the thing, has no instrument that tests it, no budget line of its own, and no place on the critical path.7 min read
- Operational readiness has seven conditions. Handover delivers two.The phrase travels across an office move, a production line, a hospital wing and a finance function taking on a new reporting basis, and it survives the journey by meaning nothing specific in any of them. ISO 55001 already makes it specific, and the arithmetic that falls out is uncomfortable.7 min read
- The only test the building ever getsCommissioning is the one assurance instrument that measures instead of reads. The standards that require it price its independence in tiers, publish their reasons, and there is a 1,500-building dataset that says what the money buys. It also says less than it used to, which is worth understanding rather than hiding.11 min read
- Under FIDIC, taking over comes before acceptanceUnder FIDIC 2017, taking over and contractual acceptance of the works are separate. At taking over the owner takes possession and may begin operating; Sub-Clause 11.9 makes the Performance Certificate the document that constitutes acceptance, and it follows a Defects Notification Period of one year unless the Contract Data states otherwise. A general law-firm summary lists what can change at practical completion or taking over: the contractor's insurance of the works ends, the risk of loss transfers, exposure to delay damages stops, and, where the contract provides for it, retention and the performance bond can begin to release. Where they do, much of the hold on the contractor loosens just as the owner's exposure begins.9 min read
- A delivered file is not an accepted recordA handover contract can test that the information arrived and that the Engineer had no objection to it. That proves receipt. It does not prove that the data passed a schema, that its identifiers reconcile with the finance and maintenance records, or that the receiving system can run a transaction on it. Those are three further tests, and where a contract's information conditions end at delivery and absence of objection, each can fail while those information obligations are performed in full.13 min read
Whether the records describing what the entity owns are true, and whether the management system around them has ever changed a decision.
What to examine
- An asset management system that is scopedIs there a defined asset management system, and is it written down which assets it covers and which it does not?
- The SAMP and the line back to organisational objectivesCan any asset management objective be traced upwards to an organisational objective, and downwards to something someone is doing this year?
- Asset data and its qualityIs the data the organisation makes decisions from of a quality anyone has determined, and does anyone know when it was last true?
- Performance evaluation and internal audit of the systemIs the asset management system itself audited, by people competent to audit it, on a programme derived from risk?
Instruments
- Asset register package
The financial register, the technical register and the crosswalk between them, generated together as one Excel workbook. The two registers share the asset tag, and every column states the requirement behind it.
- Criticality screen
Find the assets that have to be looked at separately, and see what a likelihood times consequence ranking does to that list.
Related writing
- Audited on the transactions, unaudited on the basisWhere an internal audit function covers the asset base at all, it usually covers maintenance spend and procurement compliance, because those resemble the rest of the audit universe. The decision framework, the plan and the data go unexamined. The organisation ends up assured about how it spends and unassured about what it decided to spend on.6 min read
- Every organisation has a risk register. Ask for the other oneThe 2024 edition separated actions to address risk from actions to address opportunities, which the 2014 edition ran together in a single subclause. That is not a drafting tidy. When one process has to serve both, loss avoidance wins, because it is the half with a committee, a template and an auditor.9 min read
- Nothing tells you when a record stopped being trueIn the 2024 edition Clause 7.6 is data and information, revised and set beside documented information at 7.5. Controlling a record and knowing it is still true are different questions. It is the difference between holding a register and being able to defend one.8 min read
- A dashboard computed from your own beliefs is not informationThe test for any instrument on a capital programme is whether its output contains something the inputs did not. A rule applied, a calculation done, a source consulted. A rating a team gives itself contains none of these, and colouring it does not add one. One public dashboard has carried both kinds side by side since 2009, and an auditor has re-rated the self-assessed column and published the result.in Delivery10 min read
- A strategic asset management plan that has never stopped anything is a descriptionClause 6.2.1 gave the SAMP its own requirement in the 2024 edition, and ISO strengthened the leadership clause in the same revision. The two changes belong together. A plan only allocates if somebody senior enough is willing to let it decline something, and a plan that has never declined anything has not been asked to allocate.9 min read
- Conformance is not maturity, and the body that says so is selling nothingTwelve national asset management societies agreed a position statement that says conformance to ISO 55001 may not deliver the value stakeholders want. It names four levels and publishes the characteristics of a mature organisation. It says it is not itself an assessment tool, points to companion guidelines for assessing maturity, and expects each member society to write its own detailed guidance. An owner is better off knowing whose scale produced a maturity score than buying the first ladder they are offered.6 min read
In the asset lifecycle method
The steps at which the assurance family acts in the asset lifecycle method, 4 in all, by stage and seam: what it does, in which capacity, and whether the assignment is required, analysis or the organisation’s own decision. A family is a kind of work, not this role’s job description; each listed step contains the proposition on which the assignment is based.
Inception & Development
- Delivery and contracting strategyreviews; participant; required
- Sanction and the authority to stopreviews; participant; your organisation decides
Handover & Transition
- Operational acceptancereviews; participant; your organisation decides
Operations & Asset Management
- Benefits realisationreviews; participant; your organisation decides
Lifecycle review method
- Purpose, authority and independence
- Entry point and scope
- Reviews that begin partway through the lifecycle
- Selecting the domains to examine
- Criteria set before testing
- Evidence request and provenance
- Sufficiency and the reasoning behind any sample
- Testing and documentation
- Recording what the testing found
- The views of the responsible party
- Management response and disagreement
- Reporting
- Action ownership
- Follow-up and closure on fresh evidence
Where the assurance function acts in the method Lifecycle review method
This page composes material published elsewhere on this site and adds no claim of its own. The arrangement into stages is this publication’s, as the framework says at length; the requirements above are quoted from documents that are free to read and linked next to each. Where a stage is empty for this role, it says so rather than being left out.




