The UK government’s pledge to halve consultancy spending has a definition problem. A structural one that the Public Accounts Committee’s March 2026 report identifies, names, and then declines to fix.
The Chancellor announced in November 2024 that central government would halve its consultancy spend, saving over £1.2 billion by 2026. The Cabinet Office chose to benchmark this target against a figure of £1.34 billion. Other data sources put the actual spend at £2.23 billion. That is a gap of nearly £900 million. The PAC’s finding is direct: government ‘does not have a grip on levels of consultancy spend’, the Cabinet Office ‘appears unconcerned at the many and obvious inconsistencies’, and non-compliance with existing directives is ‘going unmonitored’.
The report then identifies the structural reason: ‘Departments often have large contracts that combine consultancy and professional services, making it difficult for departments to isolate the spend on consultancy services only’. And it names the consequence explicitly: ‘The variability in definition and lack of rigour in reporting consultancy spend may also allow departments to reclassify spending and under-report rather than reduce spending’.
That is not a diagnosis of a data integrity risk. It is a description of the incentive the target creates.
A spending target without a definitional boundary is not a control. It is a pressure. The rational response for a department under a consultancy spending target, with a bundled contract that mixes consultancy and professional services, is to reclassify. Not to stop using the consultants. To stop counting them as consultants. The Cabinet Office’s own data infrastructure makes this easy: the NAO found that government ‘does not collect data on how it uses consultants, only what it spends’. Without activity-level data, reclassification is not detectable. It disappears into the aggregate.
The PAC’s recommendations do not change this structure. Publishing a list of non-compliant departments and requiring three-year horizon data are both reporting requirements layered on top of contracts designed to make reclassification trivial. Better reporting requirements on bundled contracts produce better-documented reclassification. They do not produce lower spending.
The question the committee did not ask is the one that matters: what contract structure would make reclassification infeasible rather than merely visible? Unbundling requirements for contracts above a threshold would force departments to isolate what they are actually buying. Mandatory activity-based reporting at contract level, rather than aggregate departmental figures, would make reclassification a contract breach rather than a classification judgement. The NAO’s own finding points directly here: the data problem is downstream of the procurement design problem.
The Cabinet Office spent 2025 working on revised definitions while departments classified as they chose. At the time of the PAC report in March 2026, no revised definition was in place. A definition that takes more than 16 months to finalise is not a technically difficult problem. It is a decision the Cabinet Office has not made. The spending target is not waiting on better data. It is waiting on a definition that would make the target meaningful, and no one appears to be in a hurry to write one.