R&D tax relief
How exposed is your R&D claim to an HMRC enquiry?
HMRC checked 17% of R&D claims in 2023 to 2024, up from 10% the year before (HMRC's own report, Approach to R&D tax reliefs 2023 to 2024), and it doesn't publish how it picks them. It does publish where it finds the most wrong claims, and which mistakes it sees again and again. This scores your claim against that published list, feature by feature, and tells you which ones you can still do something about.
What HMRC publishes
A risk assessment you can't see, and findings you can
Every claim goes through a risk assessment when it arrives, and HMRC says the Additional Information Form now lets it target that process more accurately (HMRC's approach to R&D tax reliefs, 2023 to 2024). The rules themselves stay private. Alongside them, HMRC runs a Mandatory Random Enquiry Programme: it opens checks on a random sample of SME claims and publishes what it finds. That sample is the best public evidence there is of which claims turn out wrong, and it is what the profile half of this score is built on.
The other half comes from HMRC's Guidelines for Compliance GfC3, which set out what it expects a claimant to be able to show, and from the problems this site's R&D pages already cover: thin narratives, records reconstructed afterwards, missed notification, the form, overseas costs, and agents.
Your claim
Score your claim
Free to use, nothing is saved or sent anywhere. Runs entirely in your browser. Answer for one claim, filed or still being prepared.
Enquiry risk level
Tick what applies above
Validity problems 0 · Serious markers 0 · Evidence weaknesses 0 of 7 · Profile features 0 of 5
How the level is set, our rule rather than HMRC's: any validity problem gives "Invalid as it stands". Otherwise a serious marker, three or more evidence weaknesses, or one weakness alongside three or more profile features gives "High"; any weakness, or two or more profile features, gives "Elevated"; anything less gives "Lower". Profile figures from HMRC's random enquiry sample of SME claims received in 2021 to 2022 (annex B, read on GOV.UK on 23 September 2026). A claim size or sector counts as a profile feature when HMRC's non-compliance figure for it is above the 25% for all SME claims. Not tax advice, and not a prediction of whether HMRC will open a check.
Worked examples
Three claims, scored
Example claims, not real companies, put through the same rule as the tool above.
| Claim | Profile features | Evidence weaknesses | Serious markers | Level |
|---|---|---|---|---|
| First claim, £15,000 of spend, a shop, prepared by an agent, a generic narrative and records written up afterwards | 5 of 5 | 2 of 7 | 0 | High |
| Repeat claimant, £600,000 of spend, a manufacturer, staff time apportioned without a recorded method | 1 of 5 | 1 of 7 | 0 | Elevated |
| Repeat claimant, £1.2 million of spend, a software company, records kept as the work happened | 0 of 5 | 0 of 7 | 0 | Lower |
The first claim hits every profile feature: HMRC found 64% of the value of first claims non-compliant, 50% for claims of £10,000 to under £20,000, and 64% for wholesale and retail. None of that can be changed. The two evidence weaknesses can. The second claim is "Elevated" on a single fixable point, plus a size band (£500,000 to under £1,000,000, 28%) that sits just above the all-claims figure. The third has nothing on the list: claims of £1 million or more had 12% of their value non-compliant, the lowest of any band, and information was among the lower sectors at 22%.
What each feature rests on
The features, and where each one comes from
| Feature | Group | Source |
|---|---|---|
| Claim notification needed and not made in time | Validity | GfC3 part 2: the claim is invalid. Covered on the three-year rule and missed the window |
| No Additional Information Form | Validity | GfC3 part 2. Covered on the Additional Information Form |
| Credit paid to a nominee account | Serious | More than 90% of claims that are fraudulent or show markers of fraud use nominee bank accounts. Since April 2024 agents can't normally receive a client's credit this way (HMRC report, section 2.1) |
| Unsolicited speculative claim for high commission | Serious | In July 2023 HMRC wrote to around 7,500 companies, starting with care homes, telling them what to do if approached this way (HMRC report, section 2.3) |
| Generic technical narrative | Evidence | GfC3 calls a bare field description "too vague" (part 5). Covered on R&D tax relief and an enquiry has opened |
| No competent professional view on record | Evidence | "HMRC have reduced the value of some claims to zero when a competent professional has not identified that a company was seeking an advance" (GfC3 part 3) |
| Records reconstructed afterwards | Evidence | GfC3 calls it good practice to keep records from the start of the project (part 5). Covered on evidence and records |
| Estimates with no recorded method | Evidence | Estimates should be arrived at "using evidence and reason", with the method recorded (GfC3 part 5) |
| Project start and end not set down | Evidence | HMRC sees "common mistakes in identifying the beginning and end of an R&D project" (GfC3 part 1) |
| Nobody at the company can explain the claim | Evidence | "Getting your claim right is your responsibility, even if you use a tax advisor" (GfC3 part 1). Covered on director liability |
| Untested overseas costs, periods from 1 April 2024 | Evidence | Overseas contractor and worker costs are excluded unless conditions necessary for the R&D are present only abroad and it would be wholly unreasonable to replicate them in the UK; the cost of the R&D and the availability of workers are disregarded (CIRD150500). Covered on overseas R&D costs |
| First claim | Profile | Table B.3, below |
| Claim size band above the all-claims figure | Profile | Table B.1, below |
| Sector above the all-claims figure | Profile | Table B.2, below |
| A sector HMRC targeted with education letters | Profile | Care homes, childcare providers, personal trainers, wholesalers and retailers, pubs and restaurants, named as sectors where R&D is unlikely (HMRC report, section 2.3) |
| Specialist R&D agent | Profile | 54% of claims used one. 28% of the value of those claims was non-compliant, against 21% for claims made without an agent, which HMRC describes as "slightly higher" (annex B, theme 5, tables B.4 and B.5). Check an adviser against our register |
HMRC's random enquiry data
Non-compliance by claim size
Table B.1 of HMRC's approach to R&D tax reliefs, 2023 to 2024 (published 30 October 2024): SME scheme claims received in 2021 to 2022, by qualifying expenditure. "Non-compliant by value" is HMRC's estimate of the share of the money claimed that shouldn't have been.
| Expenditure | Claims received | Fully compliant | Wholly non-compliant | Partly non-compliant | Non-compliant by value |
|---|---|---|---|---|---|
| under £10,000 | 8,800 | 56% | 35% | 10% | 37% |
| £10,000 to under £20,000 | 8,800 | 47% | 43% | 10% | 50% |
| £20,000 to under £30,000 | 8,900 | 57% | 31% | 11% | 41% |
| £30,000 to under £40,000 | 6,700 | 29% | 41% | 30% | 51% |
| £40,000 to under £50,000 | 4,500 | 58% | 16% | 26% | 30% |
| £50,000 to under £75,000 | 7,700 | 34% | 36% | 30% | 53% |
| £75,000 to under £100,000 | 3,700 | 53% | 20% | 27% | 34% |
| £100,000 to under £250,000 | 14,300 | 44% | 33% | 23% | 43% |
| £250,000 to under £500,000 | 10,200 | 51% | 23% | 26% | 35% |
| £500,000 to under £1,000,000 | 5,800 | 53% | 17% | 29% | 28% |
| £1,000,000 or more | 4,800 | 75% | 6% | 19% | 12% |
| All SME claims | 84,300 | 49% | 30% | 21% | 25% |
Claim counts are HMRC's, rounded. The compliance percentages are for the claims HMRC sampled. These claims predate claim notification, the Additional Information Form and the merged scheme, so today's rates will be different: the same report gives an illustrative estimate of 14.6% error and fraud for SME claims in 2023 to 2024. It doesn't break that estimate down by size, sector or first claims, which is why the tool uses the random sample.
Non-compliance by sector
| Sector | Claims received | Fully compliant | Wholly non-compliant | Partly non-compliant | Non-compliant by value |
|---|---|---|---|---|---|
| Accommodation/Catering | 1,200 | 16% | 57% | 28% | 91% |
| Administration | 5,400 | 49% | 44% | 7% | 26% |
| Construction | 7,100 | 58% | 22% | 20% | 27% |
| Finance | 1,100 | 43% | 44% | 13% | 38% |
| Information | 22,200 | 56% | 26% | 17% | 22% |
| Manufacturing | 20,000 | 52% | 26% | 22% | 15% |
| Professional | 13,900 | 45% | 28% | 27% | 20% |
| Other services | 7,500 | 34% | 38% | 28% | 44% |
| Wholesale/Retail | 5,800 | 40% | 33% | 27% | 64% |
Table B.2 of the same report, sectors by the company's main SIC classification, as HMRC labels them. HMRC notes the most compliant were manufacturing and professional services.
First claims and repeat claims
| Company claiming for the first time? | Claims received | Fully compliant | Wholly non-compliant | Partly non-compliant | Non-compliant by value |
|---|---|---|---|---|---|
| Yes | 18,300 | 41% | 46% | 13% | 64% |
| No | 65,900 | 51% | 25% | 24% | 21% |
Table B.3. Repeat claimants were more often partly wrong (24% against 13%), first claimants far more often wholly wrong (46% against 25%).
How often HMRC checks
Checks are more common, and most end with an adjustment
| Compliance checks | 2022 to 2023 | 2023 to 2024 |
|---|---|---|
| Claims received | 90,000 | 61,000 |
| Compliance checks | 8,300 | 9,700 |
| Checks as a share of claims (coverage) | 10% | 17% |
| Tax recovered from checks | £288 million | £441 million |
| Average time to complete a check | 269 days | 246 days |
| Checks where an adjustment was required | 71% | 77% |
Table 4 of the same report, rounded to the nearest 100 and excluding cases dealt with by HMRC's Large Business teams. HMRC adds that 89% of the checks it settled in 2023 to 2024 were resolved by agreement with the claimant.
What to do with the result
Profile you can't change. Evidence you can
A first claim in a small band from a retailer will always look like HMRC's riskiest group on paper. What decides how an enquiry goes is whether the evidence answers the questions a letter asks: the narrative, the competent professional's view, the records and the cost method. If the claim hasn't been filed, that is where the time should go (what's worth keeping). If it has been filed and you now think it's wrong, correcting it before HMRC asks is treated differently from waiting (you think a past claim was wrong). HMRC's own R&D disclosure service warns that a company may be charged extra interest and penalties if it waits for HMRC to contact it. And if a letter has already arrived, see an HMRC enquiry has opened.
If the claim is being used to raise money, a lender will look at the same features: an open enquiry changes what an advance looks like (R&D advance funding).
Sources
- HMRC's approach to Research and Development tax reliefs 2023 to 2024 (published 30 October 2024): annex B tables B.1 to B.5, table 4, sections 2.1 to 2.3
- HMRC Guidelines for Compliance GfC3: help to see if your work qualifies as R&D for tax purposes, parts 1, 2, 3 and 5
- HMRC manual CIRD150500: overseas expenditure restrictions (updated 25 August 2026)
- GOV.UK: tell HMRC if you've claimed too much R&D tax relief
Talk it through
Need another perspective?
Describe where things stand in a sentence or two, and we'll tell you whether it's something we can help with. There's no charge for this.
Thank you. It's with our team now.
A person reads every enquiry and we'll come back to you with what we think the right next step is. No obligation at any point.
What happens next
- A person on our team reads it. A sentence or two is enough to start.
- If we can help, we may introduce you to a provider and tell you who they are.
- No charge and no obligation at any point. You decide whether to go further.
Adam Parker
Founder of Muswell Rose Consulting Ltd, which trades as Established Finance · former Managing Director of Penny, an invoice finance business, working in mortgages, commercial finance and fintech lending since 2010 (career history).
Last reviewed:
Practical questions
Before you get in touch
What makes HMRC open an enquiry into an R&D claim?
HMRC runs every claim through a risk assessment and uses the Additional Information Form to target it, and it also checks a random sample. It doesn't publish its risk rules. What it does publish is where it has found the most error: first-time claims (64% of their value non-compliant in its latest random sample, against 21% for repeat claimants, Table B.3 of HMRC's Approach to R&D tax reliefs 2023 to 2024), smaller claims, sectors where R&D is unlikely, and claims whose credit was paid to a nominee account.
How many R&D claims does HMRC check?
HMRC checked 9,700 claims in 2023 to 2024, 17% of the 61,000 received, up from 8,300 checks and 10% coverage in 2022 to 2023. Those figures exclude claims handled by its Large Business teams. Source: HMRC, Approach to R&D tax reliefs 2023 to 2024 (GOV.UK, October 2024).
Is a small R&D claim more likely to be wrong?
In HMRC's random sample of SME claims received in 2021 to 2022 (Table B.1 of its Approach to R&D tax reliefs 2023 to 2024 report), claims with expenditure of £1 million or more had 12% of their value non-compliant. Every smaller band was higher, from 28% up to 53%. A small claim isn't wrong because it's small, but it is judged against the same evidence bar.
Does this tool predict whether HMRC will open a check?
No. It counts how many features that HMRC's own reports and guidance link to wrong claims are present in yours, and turns the count into a level using a rule we state on the page. HMRC's selection method isn't public, and it also picks claims at random.
What information do I need?
To start, just a description of what’s actually happening in the business. If it progresses, the provider will ask for the usual things: recent accounts, a sense of turnover and trading history, and details of the specific need.