An engineering firm in Rotherham is cold-called by an agent who says they are almost certainly sitting on an R&D claim. The agent takes 25% of whatever comes back, so there is nothing to lose. A claim goes in for £41,000 covering a year of work designing bespoke jigs for customers.
Two years later HMRC opens an enquiry. The jigs were skilled, careful and commercially valuable, and every one of them was built using methods the firm already had. The claim is withdrawn, the money is repaid with interest, a penalty is considered, and the agent's 25% is not coming back.
The test is about the field, not about you
The statutory definition of R&D for tax purposes points to guidelines issued by the department responsible for science, originally published in 2004 and updated since. The core test is that a project seeks an advance in science or technology through the resolution of scientific or technological uncertainty.
Two words do the heavy lifting. The advance must be in the overall knowledge or capability in a field of science or technology, not merely in your own company's knowledge or capability. Doing something for the first time is not enough if a competent professional working in that field could have done it from existing knowledge. The question is not whether it was new to you. It is whether it was new.
Uncertainty means that a competent professional in the field could not readily deduce whether something was technologically feasible, or how to achieve it in practice. If the answer is knowable by someone experienced and the only problem is that you had to work at it, that is not uncertainty in the relevant sense. Deadline pressure, budget pressure and commercial risk are not technological uncertainty either.
Some other exclusions surprise people. Work in the arts, humanities and social sciences does not qualify, and neither does economics. Routine analysis, cosmetic improvements, and applying existing techniques to a new customer's problem do not qualify. Producing a bespoke version of something your industry already knows how to make almost never qualifies, however difficult it was.
Software claims are the most contested area. Building a website, an app or an integration using established frameworks is ordinary development. The claimable part, if there is one, is usually a specific technical problem within it where the existing state of the art genuinely did not provide an answer, and the boundary of the project has to be drawn around that problem rather than around the whole build.
The schemes have been rebuilt, so old advice is wrong
For accounting periods beginning on or after 1 April 2024, the separate SME scheme and the Research and Development Expenditure Credit scheme were replaced by a single merged scheme for most claimants. The merged scheme works as an above-the-line expenditure credit, which is taxable, so the headline percentage and the cash benefit are different numbers.
Alongside it sits Enhanced R&D Intensive Support for loss-making SMEs whose qualifying R&D expenditure is a high enough proportion of total expenditure. The intensity threshold was reduced for periods beginning on or after 1 April 2024, and there is a grace year so that a company which qualifies does not fall out immediately after one unusual year.
Restrictions on overseas expenditure also came in. Subcontracted R&D and externally provided workers generally only qualify where the work is undertaken in the UK, subject to narrow exemptions where the necessary conditions are not present here and it would be wholly unreasonable to replicate them. Cost, and the availability of workers, are expressly not sufficient reasons.
Because the rules changed by reference to when an accounting period begins rather than when the spend occurred, a great deal of guidance written before 2024 is now simply inaccurate. Check the date on anything you read, including this.
HMRC has changed how it looks at claims
Two procedural requirements now catch out claimants who did everything else right.
A claim notification must be submitted in advance for companies that are new claimants or that have not claimed in the preceding three years. It is due within six months of the end of the period of account. Miss it and the claim is invalid, regardless of how good the underlying R&D was. This is the single most common way a legitimate claim is lost entirely.
An additional information form must be submitted before or at the same time as the return containing the claim. It requires a breakdown of qualifying costs, details of the projects, the name of a senior officer of the company responsible for the claim, and the agent involved. Submitting the return without it means the claim is removed.
HMRC has published its own estimates of error and fraud in the reliefs and has substantially increased compliance activity, including a mandatory random enquiry programme that selects claims without any suspicion attaching to them. The practical effect is that a claim now needs to be written to be defended, with a project narrative that identifies the competent professional, states the baseline of existing knowledge, names the uncertainty, and explains what was done to resolve it.
The agent's incentive is not your incentive
A contingent fee model pays the agent a percentage of the benefit obtained and nothing if no claim is made. The rational behaviour for an agent on that model is to submit claims, because a claim that survives pays and a claim that fails costs them little. The company carries the repayment, the interest, the penalty and the enquiry.
Warning signs are consistent. An unsolicited approach, particularly by telephone. A confident assessment of whether you qualify made before anyone has spoken to your technical staff. A claim narrative written entirely by someone with no engineering or software background, using language your own people would not recognise. A promise that HMRC never checks. A fee that is a percentage with no cap and no responsibility for defending an enquiry.
Note also that the person signing the return is a director of your company. Responsibility for a claim rests with the company, not with the adviser who prepared it. A penalty for an inaccurate return is assessed against the company's behaviour, and behaviour includes whether you took reasonable care in choosing and instructing your adviser.
If you do genuine R&D, the relief is worth having and you should claim it. The evidence to do so properly is contemporaneous: project records, test results, the failed approaches as well as the successful one, and a named competent professional who can describe in their own words what was not known at the outset. A firm that keeps those records has a claim it can defend. A firm that reconstructs them two years later from timesheets has a claim it cannot.
Where this shows up in the game
Business Baron has a research tree that is specific to each trade and gated by era, and relief on it behaves the way the real relief behaves. There was nothing at all before April 2000, so an early career gets no help with development costs. Projects can fail and produce nothing, which is exactly the uncertainty the legislation is trying to describe. Claim aggressively and you may draw an enquiry later. The five business types in the free game have their trees too.