Carillion made contract accounting and public-service continuity an outsourcing accountability test

Summary

• The liquidation was a trigger, not a complete root cause. The immediate event was Carillion's inability to secure enough support to continue and the making of compulsory winding-up orders on January 15, 2018. The deeper accountability chain ran through contract estimates, financial reporting, debt and cash pressure, board information, audit challenge, supplier dependence and the government's exposure to one strategic supplier.

• Long-term contract accounting converted uncertain future outcomes into current reported performance. That is not inherently improper. It becomes dangerous when project teams, divisional management, finance, directors and auditors cannot reconcile forecast cost, recoverable claims, certified revenue, cash collection and downside exposure. Later FCA and FRC outcomes identified serious failures within their respective legal and professional scopes.

• Profit and liquidity were not interchangeable. Reported margin could rise before a disputed claim produced cash, while slow customer receipts, debt, pension contributions and long supplier terms still consumed financial capacity. Reverse factoring could accelerate payment for a supplier while extending the period before Carillion reimbursed the financing bank. A board needed a single view of those obligations rather than separate optimistic measures.

• Detection failed at several layers. Project-level deterioration, contract-risk information and cash signals did not consistently become decision-grade information for the board and audit committee. External audit did not provide the challenge later enforcement records said was required. Government monitored Carillion as a strategic supplier, but the July 2017 warning still contradicted the picture available to officials.

• The public response limited one class of harm while leaving others. Government-funded trading liquidation and contract transfers preserved many operational services and jobs. Some construction work nevertheless stopped, suppliers and unsecured creditors faced losses, employees were made redundant, pension schemes entered protection arrangements, and public money funded continuity and liquidation activity.

• Legal outcomes must stay separated. Parliamentary committees made investigative and policy findings. The winding-up orders were insolvency procedure, not a court finding on contract accounting, audit or director conduct. The FCA made final market-disclosure and control findings against defined legal subjects. The FRC imposed audit and accountancy sanctions under professional schemes. Director disqualification undertakings had their own statutory effect. The Pensions Regulator concluded that the evidence did not permit specified anti-avoidance or criminal powers. None is a universal finding of fraud or criminal liability.

• Reform is proved by operation, not publication. Later sourcing guidance added should-cost analysis, financial-standing assessment, risk allocation, resolution planning and continuity requirements. Those controls are useful only if departments and suppliers maintain current data, test transfer plans, expose deteriorating contract economics and show that small firms and service users can withstand a supplier failure.

Scope and evidentiary boundaries

This analysis concerns Carillion plc's collapse as an accountability test for contract accounting and continuity of outsourced public services. It does not attempt to reconstruct every subsidiary, foreign project or private dispute. It separates Carillion plc from group companies, directors, auditors, advisers, contracting authorities, the Official Receiver, pension bodies and regulators because each controlled a different part of the system and each formal outcome applied a different evidentiary standard.

The starting institutional record is the UK Parliament Carillion joint inquiry publications gateway. The joint inquiry assembled oral evidence, company and adviser correspondence, pension material and a committee report. Its conclusions are important legislative findings, but they are not findings by a civil or criminal court. Strong committee language is attributed here rather than silently converted into an adjudication.

The full Business, Energy and Industrial Strategy and Work and Pensions Committees report supplies a detailed chronology and analysis of contracts, debt, dividends, pensions, suppliers, governance, audit and government. It also reproduces or characterizes evidence from witnesses and documents. Where witnesses disagreed, the existence of testimony is confirmed; the underlying disputed proposition is not treated as proven merely because it appears in the report.

Corporate filing provenance is checked against Carillion plc's Companies House filing history. The register confirms when accounts and corporate documents were filed, including the group accounts made up to December 31, 2016 and the later winding-up order. Companies House itself warns that it does not verify the accuracy of filed information. A filing therefore proves submission and public availability, not the truth of every estimate within it.

Government handling and continuity are assessed primarily through the National Audit Office investigation. The NAO examined strategic-supplier monitoring, contingency planning, the request for support, liquidation response and public cost. It expressly did not assess management of the pension schemes or every action of directors and advisers. Its dated cost estimates are not final liquidation accounts.

Later legal precision comes from the FCA's final enforcement announcement concerning two former finance directors and the FRC's 2023 announcement of audit sanctions. Those records supersede the temptation to treat early allegations as settled outcomes. The FCA instruments address market announcements, listing obligations and the conduct of named subjects. The FRC instruments address professional and audit requirements. Neither source, by itself, decides the full causal history of the liquidation.

Three vocabulary rules govern the analysis. Confirmed means supported by a filing, final notice, official liquidation record or clearly attributed official investigation. Inference means a reasoned conclusion from those records, stated as such. Disputed means the public record contains competing accounts or a committee characterization not adopted by an adjudicator. Unknown means the reviewed public evidence does not establish the answer. These labels prevent the size of the collapse from doing the work that evidence must do.

The control system before the warning

Carillion combined construction, facilities management and public-service contracts across a large group. That mix created two different risk clocks. Facilities services could generate recurring cash tied to continuing performance. Construction contracts could run for years, with current profit depending on forecasts of remaining cost, approved variations, claims, delays and completion. A group-level result could therefore look stable while a small number of large projects accumulated losses that had not yet been fully recognized.

Long-term contract accounting necessarily uses estimates. Management must forecast total revenue and total cost, then recognize performance as work progresses under the applicable standard. Estimates change as design, productivity, materials, subcontractor performance, customer instructions, delay and claims evolve. The existence of judgment is not evidence of misconduct. The control question is whether judgments were supported by current project evidence, challenged independently, reconciled to cash and revised promptly when evidence deteriorated.

Four figures should have been connected for each material contract. The first was the operational team's estimate of cost to complete. The second was the amount finance expected to recognize as revenue, including any claim or variation. The third was certified or otherwise contractually supportable entitlement. The fourth was cash collected. If forecast profit depended on a claim that was not agreed and not collected, the board needed to see both the accounting estimate and the liquidity exposure. A positive margin without cash conversion could not fund debt service, pension contributions or suppliers.

The joint committee described a business model that relied on acquisitions, debt and winning new work. That is a parliamentary assessment, not a statutory cause finding. The underlying risk mechanism is nevertheless testable. New contracts can provide mobilization cash and headline revenue while older projects require cash to cover delay or disputed recovery. Growth then masks deterioration only while the organization continues to win work and obtain financing. Once confidence falls, the inflow that sustained the model can contract faster than long-term obligations can be reduced.

Debt statistics also needed a wider perimeter than year-end net debt. A year-end snapshot can be affected by collection timing, delayed payments, disposals and short-term facilities. Average borrowing, peak borrowing, supplier days, reverse-factoring balances, pension payments and contract receivables reveal different aspects of financial capacity. The accountability failure is not established by one high number. It arises when decision makers cannot see how those measures interact under stress.

Supplier finance illustrates the distinction. Under Carillion's early-payment facility, a participating supplier could receive money from a bank before the standard payment date, less a financing cost, while Carillion reimbursed the bank later. Such arrangements can improve a supplier's immediate liquidity. They can also extend the buyer's effective financing and make trade-related obligations look different from conventional borrowing. The proper control was transparent classification, maturity reporting and stress testing, not an assumption that the facility was either inherently benign or inherently abusive.

Dividend and pension decisions belonged in the same capacity view. The parliamentary report said Carillion paid GBP 441 million in dividends during 2011 through 2016 and GBP 246 million in pension deficit-recovery payments, and it criticized the final 2016 dividend paid in June 2017. Those are committee findings from a defined period. They do not establish that every pound retained would have gone to pension schemes or prevented insolvency. They do show why a board considering distributions needed a downside case based on contract losses, debt, liquidity and obligations rather than reported earnings alone.

Detection: information existed without becoming a reliable stop signal

A large contractor normally produces abundant information: project forecasts, cost-value reconciliations, claims registers, cash reports, risk logs, internal audit work, external audit evidence and board papers. Quantity was not the central problem. The problem was whether adverse project evidence could change the group forecast, constrain a market statement, stop a dividend, reduce a bid, trigger an equity discussion or alter financing plans.

The later FCA record gives that issue legal specificity for defined periods and people. The Richard Adam final notice states that his Upper Tribunal reference was withdrawn and that the notice contains the Authority's findings, not judicial findings. It describes project-level hard risks, contentious amounts and major divergences that were not adequately reflected in board, audit committee and market information. The notice's findings concern Mr Adam's relevant period and responsibilities; they are not a license to infer the knowledge of every director or employee.

The Zafar Khan final notice addresses a different tenure and record. Keeping the two notices separate matters because responsibility follows actual office, information and authority. A finance director can control financial-reporting procedures and escalation without controlling every operational cause of a construction loss. Conversely, a project team can identify deterioration without possessing authority to change a group announcement. Accountability depends on how that information crosses the boundary.

The FCA later issued a distinct final enforcement announcement concerning former chief executive Richard Howson. It said he withdrew his challenge and recorded findings within the FCA's market-abuse and listing jurisdiction. That outcome should neither be merged with the finance-director notices nor enlarged into a criminal conclusion. It does demonstrate that construction expertise at board level carried an information responsibility different from, but connected to, the finance function.

The institutional lesson is that a risk report must preserve disagreement. If a project team forecasts a loss but a group forecast retains a profit, the board should see both values, the reasons for the difference, the evidence required for recovery, the owner of the judgment and its cash consequence. Replacing the project number with a single approved group number destroys information at the point where challenge is most valuable.

Escalation also needs a consequence. A red flag that appears month after month without changing authority becomes descriptive rather than preventive. Material contract variance should trigger an independent review, a restriction on recognizing additional claim value, direct audit committee visibility and a liquidity adjustment until the difference is resolved. Management may ultimately reject the lower estimate, but the record should show why, who approved the decision and what later evidence would reopen it.

Internal audit and external audit had different roles. Internal audit could test contract-control design, adherence and escalation. External audit sought reasonable assurance over the financial statements and had to obtain sufficient appropriate evidence for material judgments. Neither could take over the board's responsibility to prepare accounts or management's responsibility to operate controls. The failure of one defense did not erase the responsibility of the others.

July 2017 converted accumulated uncertainty into a visible crisis

On July 10, 2017, Carillion announced an expected provision of GBP 845 million following a contract review, suspended dividends and changed leadership. The later FRC audit-sanctions announcement records that provisions announced in July and September totalled GBP 1.045 billion, primarily from expected contract losses, alongside a GBP 134 million goodwill impairment. These are different dated measures. They should not be presented as one loss discovered on one day.

The July warning mattered because it altered the credibility of prior accounts, the market's view of financial capacity and the government's assessment of a strategic supplier. According to the NAO, the scale surprised government because it contradicted market expectations and information previously supplied by Carillion. The Cabinet Office moved the supplier risk rating from amber to red and began contingency planning. The warning therefore operated as both a corporate disclosure event and a public-service continuity signal.

Profit-warning chronology should not be confused with physical project chronology. A project may have experienced delays, claims or cost deterioration before finance changed the recognized estimate. A later provision does not establish that the final amount was knowable at every earlier date. It does, however, require investigators to ask which components were known, probable or disputed at each reporting point, and whether controls updated the estimate as evidence changed.

After July, Carillion pursued disposals, cost reduction, financing and restructuring. The fact that rescue efforts continued does not prove that they were viable. It shows that the company, lenders, advisers and government faced changing options. A credible recovery plan required cash by date, not only eventual accounting benefits. It also had to account for the operational effect of reducing employees, stretching suppliers or selling assets that generated future earnings.

Government faced a different decision. Supporting Carillion could preserve continuity but transfer open-ended corporate risk to the state. Refusing support could allow liquidation but require public funding for essential services and special management. The NAO reported that Carillion sought GBP 223 million of government support in early January 2018, alongside help with restructuring. The Cabinet Office rejected the request because of concerns about the plan, legal exposure, precedent and potentially continuing calls for money. That decision addressed public risk; it did not itself cause the contract losses already embedded in the group.

The immediate trigger was the failure to secure sufficient support and liquidity, followed by compulsory liquidation. It was not a routine administration in which directors retained control while a plan was developed. The court appointed the Official Receiver, and special managers assisted a trading liquidation. Corporate control over the service portfolio changed at the moment continuity risk was highest.

That court step should not be asked to carry more evidentiary weight than it can bear. It established the insolvency process, the appointment structure and the need to manage ongoing contracts. It did not decide whether a particular contract estimate was improper, whether an audit procedure failed a professional standard, whether a director breached a market rule or how much any creditor would ultimately recover. Those questions later moved through regulators, parliamentary committees, disqualification undertakings, professional settlements and liquidation administration under different tests.

Cause classification: trigger, root conditions and consequences

The triggering event was the making of winding-up orders on January 15, 2018 after rescue and financing efforts did not produce a viable funded solution. That event made the corporate failure legally operative and transferred control to the liquidator. Calling the winding-up order the root cause would be circular: the order responded to insolvency; it did not explain how financial capacity had been exhausted.

The root accountability problem was the inability of the governance system to convert contract-level uncertainty into timely, reliable group decisions about profit, cash, debt, distributions and risk. This statement is an analytical conclusion from the official record, not a new legal finding. It places contract estimates at the center without pretending that accounting entries alone created physical project losses or liquidity needs.

The contributing conditions included debt dependence, acquisitions and goodwill, cash pressure, uneven contract performance, reliance on claims and recoveries, long supplier terms, pension obligations, dividend choices, governance incentives and the public sector's concentration on a strategic supplier. These conditions interacted. None should be declared sufficient by itself. A company may carry debt, use reverse factoring, pay dividends or hold goodwill without failing; the danger lay in the combined loss-absorption capacity when core contract expectations changed.

The detection failures occurred where project evidence did not reliably alter group reporting, where board and audit committee information lacked the adverse detail needed for challenge, where external audit did not obtain or act on sufficient evidence, and where government monitoring did not produce an early view of the scale that appeared in July. Detection is not the same as prediction. The record does not establish that one earlier meeting or report certainly would have prevented failure.

The response failures before liquidation included the limited time in which restructuring plans had to become funded and the inability to restore confidence after successive warnings. It would overstate the evidence to assign every failed negotiation to one actor. The defensible conclusion is narrower: by the final week, the proposed plan did not satisfy those asked to finance or support it, and the government judged open-ended rescue less defensible than trading liquidation.

The recovery outcome was mixed. Operational public services were largely preserved through continued funding and transfer, which was a significant continuity achievement. Construction projects did not all continue, creditors did not all receive what they were owed, and many workers and pension members bore consequences. Recovery must therefore be measured by service, contract, person and date rather than summarized as either seamless success or total breakdown.

Audit accountability: later findings replaced early speculation

The 2023 FRC outcome provides the strongest public professional record on the statutory audits for 2014 through 2016 and additional 2017 work. The regulator imposed sanctions on KPMG entities and former partners, declared that relevant audit reports did not satisfy requirements, and described failures to obtain sufficient appropriate evidence and apply professional scepticism. It also addressed going concern, debt, supply-chain finance, pensions, goodwill and major contracts.

Those findings require precision. The FRC said the relevant audit breaches were not dishonest and, in most instances, were not intentional, deliberate or reckless. It separately identified specified integrity, objectivity and independence failures. A responsible account therefore does not use the scale of the sanctions to label every audit failure fraudulent. It states the admitted or determined breach and its scope.

A separate 2022 FRC tribunal outcome concerning information supplied during audit-quality reviews involved false and misleading information and documents provided to the regulator in relation to reviews of two audits, one of them Carillion's 2016 audit. That was a professional disciplinary result about regulatory inspection evidence. It should not be conflated with every substantive judgment in the financial statements or with the later 2023 settlement concerning the statutory audits themselves.

In May 2026, the FRC announced accountancy-scheme findings and sanctions involving two former group finance directors and three other accountants. The announcement states that the named former finance directors accepted misconduct concerning several business areas, contracts and a supply-chain finance facility, and that the subjects admitted reckless conduct and failures of integrity but not dishonesty or deliberateness. Its express subject limitation must be preserved: it is not a finding against people or entities outside those proceedings.

The audit lesson is operational. Contract testing should begin with the source estimate, not the number already aggregated for group reporting. Auditors need project evidence, customer correspondence, variation status, cost-to-complete support, post-balance-sheet performance and a cash reconciliation. They should test management bias across contracts, not only whether any single estimate lies within a broad plausible range. A sequence of optimistic judgments can be material even when each one is defended separately.

Audit committees also need an unresolved-items register. It should identify every material contract where project and group views diverge, the amount in dispute, the evidence outstanding, the responsible executive, the auditor's position and the date of the next decision. Meeting minutes should show the challenge and answer. An unqualified opinion cannot substitute for that governance record, and a committee cannot delegate its understanding of major estimates entirely to the auditor.

Market-disclosure and director outcomes stayed legally distinct

The FCA's 2026 final notice for Carillion plc in liquidation records a public censure rather than a financial penalty because of the company's financial circumstances. It concerns specified announcements, market-abuse and listing provisions, attributed knowledge and the company's systems and controls. It is a final regulatory instrument, but it is not a criminal conviction and does not decide claims outside the Authority's jurisdiction.

The former finance directors' final notices and the former chief executive's outcome concern their own relevant periods and responsibilities. They should not be collapsed into a statement that every director had identical knowledge. Nor should later final findings be projected backward as if every person outside the defined periods had access to the same record. Legal accountability becomes more accurate, not weaker, when the subject, conduct, date and test remain visible.

The same separation applies to settlements and undertakings. An FCA penalty follows market-disclosure and listing-rule tests. An FRC accountancy settlement follows professional misconduct standards and can record admissions of recklessness and integrity failure without an admission of dishonesty. A director-disqualification undertaking restricts future conduct without a contested trial. A liquidation update reports administration and recovery status. Treating those documents as interchangeable would overstate some findings and erase others.

Director disqualification created another track. The Insolvency Service announced a 12-and-a-half-year disqualification undertaking by former finance director Richard Adam. The release describes the conduct accepted for that statutory purpose. An undertaking has legal effect without a contested trial, but it should be characterized as an undertaking, not a criminal sentence or a judicial finding after evidence was tried.

The Service separately announced an eight-year disqualification undertaking by former chief executive Richard Howson. It followed an investigation and avoided the need to determine the described conduct at trial. The press notice's detailed account belongs to that disqualification process. It does not automatically resolve every professional, regulatory or damages question involving the same historical facts.

These outcomes also show why the word fraud requires discipline. The official record contains findings of misleading announcements, market-abuse contraventions, reckless conduct, professional misconduct and audit breaches. Some records expressly disclaim dishonesty; some use objective or recklessness tests. This article does not replace those terms with an unsupported allegation of criminal fraud.

Pensions: impact and enforcement were not the same question

Pension figures in the public record use different bases. The NAO cited GBP 2.6 billion of pension liabilities at June 30, 2017. The joint committee discussed deficits on scheme, recovery-plan and Pension Protection Fund bases. The Pensions Regulator later described an estimated buy-out deficit of about GBP 1.8 billion at the end of 2016 across 13 defined benefit schemes. These values should not be added or treated as contradictory without matching valuation basis and date.

The Pensions Regulator's intervention report is especially important because it limits overclaiming. The regulator examined disposals, dividends, misleading information and possible anti-avoidance and criminal powers. It concluded that it could not use the specified powers, found no evidence that the disposals caused material detriment or were undertaken mainly to avoid scheme debts, and said the false-or-misleading-information criminal offence was not made out on the facts it assessed.

That result does not mean pension members suffered no impact. The regulator said the insolvency significantly affected schemes and benefits. It means impact alone did not satisfy the legal tests for those particular powers against an available target. Accountability analysis must hold both propositions at once: pension risk was material, and the regulator concluded that the reviewed evidence did not support the enforcement routes it considered.

The counterfactual about dividends is likewise bounded. Retaining distributions would have preserved some cash, but public evidence does not prove where that cash would have gone or how long it would have postponed failure. The regulator reasoned that debt pressure and the scale of contract losses affected that analysis. A defensible policy lesson is prospective: before distributions, boards should test solvency, contract downside, debt maturities and pension recovery under a common stress scenario.

Government monitoring: dependency changed the meaning of supplier risk

Carillion was not merely a vendor that could be replaced after a missed delivery. It held hundreds of contracts across hospitals, schools, prisons, transport and other infrastructure. Government therefore faced correlated exposure: one corporate failure could affect multiple departments, joint ventures, local bodies, workers and subcontractors at the same time. Contract-by-contract performance reporting could miss that aggregate dependency.

The NAO said the Cabinet Office had rated Carillion green or amber from the creation of its strategic-supplier risk system in 2011 until the July warning, with amber status from February 2016 because of performance concerns. After the warning, officials moved the rating to red and began contingency work. The report also said government did not immediately apply its highest risk category, in part because it accepted that sensitive information was already being received and did not want to precipitate collapse. That is a documented decision context, not proof that a different label would have saved the company.

The government continued with or announced substantial work after the warning. Some awards or approvals predated July, some contracts were signed later, and joint-venture partners carried takeover obligations. A simple total of post-warning awards can therefore misstate the decision sequence. The correct test is contract-specific: what had already been committed, what termination or re-procurement would cost, what partner protections existed, what new exposure was created and which official approved the residual risk.

The Public Administration and Constitutional Affairs Committee's After Carillion report widened the inquiry from one company to outsourcing design. It criticized thin make-or-buy evidence, aggressive risk transfer, weak data and limited public evidence understanding of contractor health and supply chains. Those are committee policy findings. They do not prove that every outsourced service is inferior or that a specific procurement term caused Carillion's liquidation.

The committees' recommendations also received a formal government response. A response records what government accepted, rejected or said it was already doing. It is not evidence that each promised control subsequently operated. Implementation requires dated policy, departmental use, exceptions, assurance results and outcomes under stress.

Strategic-supplier monitoring should connect six views: aggregate public revenue, contract margin and cash, debt and financing, pension and other fixed obligations, supply-chain dependence, and a tested resolution plan. It should also preserve the distinction between confidential financial information and verified information. Receiving a document is not the same as testing its assumptions or reconciling it with contract performance across departments.

Compulsory liquidation tested continuity in real time

The government's January 15 information notice identified the companies initially subject to winding-up orders, the appointment of the Official Receiver and special managers, employee and creditor routes, and the intention to continue operating public services. It also referred to around 450 government contracts and 38 percent of 2016 reported revenue. The NAO later used around 420 public-sector contracts. The difference illustrates why counts must retain their source, date and scope.

Trading liquidation created a controlled bridge between corporate failure and replacement provision. Government funded the cost of continuing relevant public contracts while the Official Receiver and special managers identified customers, staff, suppliers, data, assets and counterparties. That avoided an automatic shutdown of every service on the morning of liquidation. It did not restore Carillion as a solvent enterprise or guarantee payment of pre-liquidation claims.

Early workforce outcomes moved quickly. A February Official Receiver employment update reported 919 jobs safeguarded through transfers and 377 redundancies at that point. Those numbers were snapshots, not final totals. They show why an aggregate announcement on liquidation day could not measure the eventual employment outcome.

By August, the Insolvency Service said agreements were in place to transfer the last of 278 contracts. It reported 13,945 jobs saved, 2,787 redundancies and continued delivery of essential public services without a major service incident during the trading period. It also said work remained on transitional services and supplier accounts. The 278 transfer count should not be subtracted mechanically from the earlier 420 or 450 figures because the categories and treatment of contracts differed.

Operational continuity did not mean every project continued. The NAO identified stoppage at two private-finance hospital construction projects while reporting that almost all services continued. A hospital cleaning or maintenance service and the construction of a hospital are different continuity units. One can continue while the other pauses. Recovery evidence should therefore identify the service, project, workforce, new provider and date rather than rely on a group-wide label.

The latest official guidance for employees, subcontractors, creditors and suppliers was updated in September 2025 with information about anticipated distributions and closure of group companies. It warns that distributions depend on asset realization and costs and cannot be treated as guaranteed. That continuing uncertainty is a recovery fact: public-service transfer was substantially faster than final creditor resolution.

The Insolvency Service's 2020-21 annual report described an estimated total deficiency of GBP 3 billion and ongoing liquidation and director work. That figure had a different scope from the NAO's estimated GBP 148 million government cost. One concerned group deficiency; the other concerned a dated estimate of government loss and liquidation support. Adding them would double-count or mix unlike measures.

Who bore the impact

Employees experienced several outcomes: transfer to a replacement provider, continued temporary employment during trading liquidation, resignation for other work, retirement or redundancy. A single number such as jobs saved cannot describe changes in pay, location, role or long-term security, and the reviewed public record does not provide a uniform outcome for every worker. The confirmed point is that continuity work preserved many jobs while thousands of redundancies still occurred.

Subcontractors and suppliers faced a different loss mechanism. Work already delivered could become an unsecured pre-liquidation claim. Continuing work might be paid under new liquidation arrangements, but that did not automatically cure old debt. Small firms had less capacity to absorb delayed or written-down receivables, and their own workers and sub-suppliers could bear second-order effects. The final number of business failures caused specifically by Carillion is not established in the reviewed source set, so no such total is asserted.

Public-service users were protected where departments, the liquidator, employees, joint-venture partners and replacement contractors kept services running. Their exposure was not limited to immediate interruption. They also depended on the quality of transferred records, maintenance schedules, staffing, security access, asset information and subcontractor relationships. A service that remains open can still accumulate transition risk if knowledge is lost.

Pension members faced movement into statutory protection and outcomes determined by scheme status and compensation rules. Creditors faced uncertain recoveries. Shareholders lost their investment. Lenders and sureties bore contractual outcomes according to their instruments. Taxpayers funded continuity and liquidation support. These impacts overlap, but they are not interchangeable and should not be combined into one headline loss.

The cost of unfinished construction also requires project-level analysis. Delay can increase financing, remobilization, redesign and deterioration costs, but the amount attributable to Carillion depends on each contract, joint-venture allocation and replacement plan. Public reports show disruption and specific paused projects; they do not support one comprehensive current total for every affected public work.

Control ownership across the accountability chain

Project management controlled current cost forecasts, productivity evidence, subcontractor status, claim support and operational escalation. Its preventive capacity was greatest before a loss became embedded. It could not alone decide group reporting or financing.

Divisional and group finance controlled consolidation, accounting policy, challenge, provisions, cash forecasting and information supplied to senior management. Finance had to preserve adverse operational views rather than normalize them into a target. It also had to reconcile reported earnings with debt, supplier-finance maturity and cash.

Executive management and the board controlled strategy, risk appetite, dividends, financing, major bids, market statements and the authority of challenge functions. Directors did not need to estimate every construction task themselves. They did need evidence sufficient to understand why project and group forecasts differed and what would happen if recovery assumptions failed.

The audit committee and external auditor controlled different challenge gates. The committee oversaw reporting and audit and could demand direct contract evidence. The auditor controlled its opinion and professional procedures. Neither could rely on the other as proof that management estimates were sound.

Contracting authorities and the Cabinet Office controlled procurement decisions, strategic-supplier monitoring, cross-government aggregation and continuity plans. They could not manage Carillion's books. They could require information, test dependency, structure joint-venture and step-in rights, assess economic standing and prepare alternatives.

The Official Receiver, special managers, departments, employees and replacement providers controlled consequence limitation after liquidation. Their work preserved services and transferred contracts. Recovery control did not retroactively prevent losses, but it reduced the risk that corporate insolvency would become immediate public-service failure.

Regulators and the courts applied powers after evidence and legal thresholds. Their outcomes create accountability and deterrence, but they do not become one pooled verdict. The absence of one sanction does not negate another authority's finding, and one final notice does not prove a proposition outside its subject and law.

Counterfactual tests

An earlier contract write-down would probably have reduced reported profit and could have changed financing, dividends, bids and market confidence sooner. It might also have caused an earlier crisis. The evidence supports the importance of timely recognition; it does not prove that earlier recognition would have produced a solvent rescue. The right counterfactual asks what options were still available at each date, not whether transparency would have guaranteed survival.

An earlier equity raise could have increased loss-absorption capacity. Whether sufficient capital was realistically available, on acceptable terms and before confidence fell is unknown from the reviewed public record. A board should nevertheless preserve that option by confronting deterioration before repeated warnings make a raise more expensive or impossible.

Stopping dividends would have retained cash. It would not have corrected contract economics, removed debt or automatically funded pension schemes. The decision still mattered because distributions reduced available liquidity and communicated confidence. A stress-based distribution gate would have forced the board to reconcile that signal with downside risk.

Stronger audit challenge could have changed recognition, disclosure or the timing of intervention. It could not itself finish a delayed hospital or collect a disputed claim. Audit was one defense in a larger operating system. Its value lay in forcing reliable information into public reporting and governance before liquidity closed the remaining options.

Moving Carillion to the highest supplier-risk category sooner might have accelerated contingency planning or constrained awards. It might also have affected confidence and contracting choices. The NAO record does not establish the outcome of that alternative. The actionable point is that risk categories need defined consequences, independent verification and decision records; otherwise changing a label does little.

What durable reform would require

1. A contract-estimate evidence chain

Every material long-term contract should maintain a versioned record of original bid assumptions, approved scope changes, cost to complete, claims, customer certification, cash, provisions and forecast range. Project, divisional and group views should remain visible. Changes should name an owner and cite evidence. A board should be able to trace a reported margin back to current operational facts without reconstructing the trail after failure.

2. A cash and obligation perimeter wider than net debt

Management reporting should show average and peak borrowing, facility headroom, supplier-finance balances and maturities, payment days, pension contributions, tax deferrals, bonding and guarantee calls, expected contract cash and concentration by customer. Stress tests should combine project losses with slower collection and loss of new work. Year-end presentation should not hide dependence on intra-period financing.

3. Automatic consequences for forecast divergence

When a project estimate and group estimate differ beyond a defined amount, the difference should reach an independent contract review and the audit committee. Recognition of additional upside should pause unless specified evidence is available. The exception should remain open until resolved, with age, owner and cash effect visible. Persistent red flags should reduce authority, not become routine narrative.

4. Audit work that tests bias across the portfolio

Auditors should sample both large contracts and patterns: early claim recognition, repeated late changes, projects reporting margin while consuming cash, estimates clustered near target, and manual adjustments at reporting dates. Engagement quality review should occur before the opinion when the risk is highest. Evidence completion after signing cannot support a decision already made.

5. Distribution and pension decisions under the same downside case

Boards should test dividends against plausible contract loss, debt maturity, supplier-payment and pension scenarios. The analysis should state which liabilities are legally due, which are discretionary and how long liquidity remains if assumptions fail. This does not make pension funding senior to every other claim by assertion; it makes the allocation decision explicit and reviewable.

6. Cross-government dependency mapping

Government should aggregate exposure to a strategic supplier across departments, joint ventures, local bodies and critical subcontractors. The map should identify which services cannot tolerate interruption, which data and assets are needed for transfer, which alternative providers have capacity and what contract rights permit intervention. A supplier's share price or published profit is not a substitute for that operational map.

7. Resolution plans tested before distress

The Cabinet Office's 2019 Outsourcing Playbook launch introduced measures including risk allocation, pilots, public performance data and supplier resolution plans. These were post-collapse reforms, not evidence of the pre-2018 standard or proof of subsequent compliance. Their value depends on whether plans contain usable service, employee, asset, data, subcontractor and financial information and whether departments test transfer exercises.

The current Sourcing Playbook carries forward financial-standing assessment, should-cost modelling, risk allocation and resolution planning. It says critical service contracts require resolution information throughout contract life. Publication establishes policy. Durable proof would include departmental assurance results, exceptions, exercise findings, time to transfer, service performance during failure and correction of stale plans.

8. Supplier continuity as a first-class service control

Critical suppliers should identify subcontractors whose failure would interrupt delivery, report payment performance and preserve direct contact and data needed for transfer. Contracting authorities should distinguish pre-insolvency debt from authorized continuation work and communicate payment terms promptly. Small firms should not have to finance public continuity through uncertainty over whether work will be paid.

9. Board information that preserves bad news

Board packs should include adverse project estimates, claim-aging, cash conversion, exceptions and prior forecast error, not only an approved point estimate. Audit committee chairs should have direct access to project finance and internal audit when a material divergence persists. Incentives should reflect cash, forecast accuracy, supplier treatment and remediation, not only revenue and reported margin.

10. Public evidence of whether reform works

Useful metrics include forecast error by contract stage, cash conversion, age of unagreed claims, supplier-payment distribution, resolution-plan test results, time to transfer a failed contract, service incidents during transition, audit exceptions and repeat findings. A policy's existence is an input. Reduced surprise, earlier escalation and bounded continuity loss are outcomes.

What remains disputed or unknown as of July 17, 2026

The public record does not provide every contract's contemporaneous estimate history, every board discussion or every lender's decision model. It therefore cannot establish one moment when failure became inevitable. Later regulators identified conduct within defined periods, while the Pensions Regulator reached a no-action conclusion under different powers. Those outcomes should remain side by side rather than forced into a single moral verdict.

Final creditor distributions and the full cost of liquidation remain dependent on realizations, claims and expenses. The September 2025 official guidance provides expectations, not guaranteed recoveries. Dated figures for taxpayer cost, group deficiency, pension liabilities and project losses answer different questions. No valid total emerges from adding them.

It is also unknown from these sources how consistently every current contracting authority applies the Sourcing Playbook, how often resolution plans are exercised, or whether supplier financial data have prevented a comparable surprise. Policy updates show institutional response. They do not by themselves demonstrate operating effectiveness.

The causal weight of each contract loss remains contract-specific. Customer disputes, project execution, bidding, design, supply chains, local conditions and accounting judgments can all matter. FCA and FRC findings establish serious reporting, control, professional and audit failures within their scopes; they do not turn every commercial dispute into proof of wrongdoing.

Finally, no reviewed evidence supports a blanket allegation that Carillion's failure was a criminal fraud. Final records use precise terms such as reckless conduct, misleading statements, market-abuse contraventions, professional misconduct, audit breaches and disqualification undertakings. Where a regulator expressly found that a criminal test was not met, that boundary is part of the evidence, not an inconvenience to the narrative.

Conclusion

Carillion's collapse became an outsourcing accountability test because uncertain contract outcomes passed through accounting, cash, debt, supplier finance, board reporting, audit and public procurement before the state had to manage the consequences. The winding-up order was the trigger. The deeper failure was a control system that did not turn deteriorating project evidence into sufficiently early and reliable decisions about recognition, liquidity, distributions, disclosure and dependency.

The response also resists a simple verdict. Trading liquidation preserved many essential services and transferred contracts, while construction projects, workers, suppliers, pension members, creditors and taxpayers still bore substantial and differently measured costs. Regulators and professional bodies later imposed consequential outcomes, but each instrument answered its own legal question. Parliamentary condemnation, FCA final notices, FRC sanctions, disqualification undertakings and a pensions no-action conclusion cannot be merged without distorting them.

The durable standard is demonstrable control before the next failure: contract estimates linked to evidence and cash, adverse views preserved for boards, audit challenge completed before sign-off, supplier obligations visible beside debt, government exposure aggregated across services, and resolution plans tested while transfer remains optional. Outsourcing does not remove public responsibility for continuity. It changes where that responsibility must be observed, tested and enforced.