Why No HMRC Enquiry Does Not Mean You Qualified

HMRC opened compliance checks into 9,700 R&D claims in 2023 to 2024, around 17% of the claims made that year. Three years earlier it had roughly 100 people working on R&D compliance across the whole of the scheme, against more than 500 by 2024. Its compliance resource was considerably smaller in that earlier period, and substantially fewer claims were subject to compliance checks than are today. That is worth holding alongside the reassurance many companies take from never having been asked about a claim.

The years that went largely unchecked are also the years HMRC’s random enquiries later found to be the worst. Its estimate for 2021 to 2022 put error and fraud in the SME scheme at 25.8% of the relief claimed. Roughly one pound in four, in a period when the chance of anyone looking at a given claim was small.

This blog is not an argument that your past claims were wrong. Most were not. It is an argument that the absence of an enquiry never told you either way, that HMRC can reach further back than most directors assume, and that the cheapest moment to find out where you stand is before HMRC picks the moment for you.

Who this applies to

This is for companies that have claimed R&D tax relief for several years without a query, and for the directors who sign those returns. It is most relevant where the claims were prepared before the Additional Information Form became mandatory in August 2023, when the evidence HMRC received with a claim was considerably thinner than it is now.

It applies with more force if any of the following is true: the adviser who prepared the claims is no longer involved, the claim was built from a conversation with a director rather than with the technical team, the same projects appeared year after year with the figures scaled up, or nobody in the business can now explain what the scientific or technological uncertainty was.

“We have always claimed and never had a problem”

This is the most common thing we hear, and it is usually said with complete sincerity. Ten years of claims, money received each time, no letters. It feels like a track record.

It is a record of not having been selected. Those are different things, and the difference only matters at the point it matters a great deal. An unexamined claim carries exactly the same risk it carried on the day it was filed, minus the years that have since run off the clock. Nothing about the passage of time makes an incorrect claim correct.

What HMRC actually did, and when it changed

HMRC’s own published account of its approach to R&D tax reliefs in 2023 to 2024 sets out the shift plainly:

• Around 100 people worked on R&D compliance in 2020 to 2021. By 2024 that was more than 500.
• Compliance checks were carried out on 9,700 claims in 2023 to 2024, around 17% of claims, up from around 10% the year before.
• HMRC identified £441 million as incorrectly claimed through those checks in 2023 to 2024, against £288 million in 2022 to 2023.

Read that backwards and the implication for older claims is the point of this blog. The 17% figure was itself a substantial increase on the previous year, and that increase followed a large expansion in HMRC’s R&D compliance resource. Many claims from the years before that were processed without HMRC ever examining their technical merits.

What the random enquiries found

The reason we know how bad the position was is that HMRC went and measured it. Its Mandatory Random Enquiry Programme takes a random sample of claims each year, examines every one, and produces a statistical estimate of error and fraud across the population. Random selection is the point: it captures what routine risk-based checking misses.

The results, as published in HMRC’s compliance approach documents and its annual report and accounts:

Year	Overall error and fraud	SME scheme
2020 to 2021	16.7% (£1,127m)	24.4%
2021 to 2022	17.6% (£1,337m)	25.8%
2023 to 2024	6.4% (£493m)	11.1% (£347m)
2024 to 2025 and 2025 to 2026	5.3% (illustrative)	Not separately published

Two things are worth drawing out. The first is the scale of the early figures. A quarter of SME scheme relief being wrongly claimed is not a fringe problem, and the claims making up that quarter were, by definition, mostly claims nobody queried. The second is that HMRC revises these estimates as new random enquiry data arrives. Its figure for 2023 to 2024 has already been restated downwards from an earlier estimate. The true position in any given year only becomes visible some years later, which is precisely the problem a company faces when it treats silence as confirmation.

Payment is not approval

HMRC’s stated service aim, set out at CIRD80525, is to pay 85% of payable tax credits within 40 days or to make contact within 40 days. Where it thinks a claim may be incorrect, it aims to open an enquiry within 60 days of receiving it.

Both are described as aims, and HMRC is explicit that they do not replace the statutory time limits for enquiring into a Corporation Tax return. A payment made inside 40 days is a processing outcome. It is not a finding that the project met the statutory definition of R&D, and it does not close anything.

What acceptance does and does not mean

Event	What it means	What it does not mean
Claim processed and paid	HMRC has made the payment	HMRC has reviewed the claim
No contact for four years	The claim was not selected	The claim was correct
Enquiry opened and closed	That period is settled on those facts	Other periods are settled
Advance assurance obtained	HMRC agreed on the facts and projects presented	Later or different projects are covered
The adviser said it qualified	You took advice	Your responsibility for the return moved

How far back HMRC can go

Two separate mechanisms matter here and they are often confused. The enquiry window is short. The discovery window is not.

The enquiry window

Where a company tax return is delivered on or before the filing date, HMRC generally has 12 months from the date it receives the return to open an enquiry. For a company that is a member of a group that is not small, the 12 months runs from the statutory filing date instead. A return delivered late, or amended, extends the window to the quarter date following the first anniversary of delivery or amendment. The detail is at EM1510.

The discovery window

Once the enquiry window has closed, HMRC can still make a discovery assessment. The time limits, set out at CH56200, run from the end of the accounting period rather than from the filing date:

• Four years where there was no careless or deliberate behaviour.
• Six years where the inaccuracy was careless.
• Twenty years where it was deliberate.

For example, a company with 31 December year ends reading this in September 2026 could still have the accounting period ended 31 December 2020 within the six-year careless assessment window, alongside each subsequent period. That window closes for the 2020 period on 31 December 2026, and for each later period a year on from that. The point is that the exposure is rarely a single claim.

What the company is actually responsible for

HMRC is direct on this point in Part 2 of its Guidelines for Compliance 3. The facts of an R&D claim always remain the company’s responsibility, even where professional advisers are involved. Its reasoning is that only the business fully understands the work it carried out.

That has consequences a lot of directors have not priced in. The repayment and the interest fall on the company. So, ordinarily, does the penalty. Whether the company has a separate commercial or contractual claim against the adviser who prepared the work is a different question, and one it would be dealing with after settling with HMRC rather than instead of it.

That is not the same as saying advisers face nothing. Where a third party deliberately supplies false information to a claimant, or withholds information from them, knowing it will produce an inaccuracy, HMRC can charge a penalty on that third party rather than on the taxpayer, under the rules at CH81165. The dishonest tax agents regime goes further, allowing penalties of between £5,000 and £50,000 on an agent, access to the agent’s working papers, and publication of their details where the penalty exceeds £5,000. Both regimes are aimed at dishonesty rather than at work that was merely poor, so neither is the answer to most weak historic claims. An adviser who knowingly manufactured one is not beyond reach.

Reliance on an adviser also affects the penalty position, even though it does not move the liability. HMRC’s guidance at CH81125 asks what information the company sought from the adviser, what evidence it was given in support, and what checks it made or could reasonably have made. A company that engaged a specialist, gave them accurate information and asked sensible questions is in a materially different position from one that signed what was put in front of it.

Where behaviour is careless rather than deliberate, CH82470 sets the penalty range at nil to 30% of the potential lost revenue for an unprompted disclosure, and 15% to 30% for a prompted one. Broadly, the distinction turns on whether the company discloses the inaccuracy before it has reason to believe HMRC has discovered, or is about to discover, it. In practical terms, identifying a problem before an HMRC intervention can therefore materially affect the penalty position.

Worked example: six years in scope

A manufacturing company with 31 December year ends has claimed R&D relief every year since 2016 and has never been contacted about a claim. Assume an average benefit received of £45,000 a year, and assume a review finds that one recurring project, representing 40% of the benefit each year, did not meet the statutory definition. Assume the behaviour is treated as careless rather than deliberate, and that the disclosure is prompted by HMRC rather than made voluntarily.

Reading this in September 2026, the six-year careless window still reaches back to the period ended 31 December 2020. The 2016 to 2019 claims are out of time on careless behaviour. Six periods remain in scope.

• Overclaimed relief: £45,000 × 40% × 6 periods = £108,000 repayable.
• Penalty: careless with prompted disclosure, so between 15% and 30% of the potential lost revenue, which is £16,200 to £32,400.
• Interest: statutory interest, calculated by reference to the relevant amounts and the dates on which they fell due. How the benefit was originally obtained matters here, because reduced Corporation Tax, a payable credit and surrendered losses do not all carry the same dates or treatment.
• Total before interest: £124,200 to £140,400.

Every figure above is illustrative and rests on the assumptions stated. What holds regardless of the arithmetic is the shape of it: the relief arrived in six instalments across six years and was long since spent, and the repayment arrives in one.

Change one assumption and the picture changes materially. Had the same company identified the problem itself and made an unprompted disclosure, the penalty range would start at nil rather than 15%, a difference of up to £16,200 on these figures before any argument about the percentage within the range.

What to do if the question is now live for you

The useful response is a review, not a decision. You cannot sensibly decide what to do about historic claims before establishing whether there is anything to do anything about, and in our experience most of the time there is not.

1. Work out which periods are actually in scope. Four years from the end of each accounting period as standard, six if carelessness is in play. That defines the size of the question.
2. Read what was submitted, not what was sold. Pull the technical narratives and the cost schedules, not the adviser’s summary or the fee note.
3. Test each project against the statutory definition, one at a time. Year-by-year conclusions are not useful. A claim is a set of projects, and they rarely stand or fall together.
4. Identify who the competent professional was for each one. If the answer is nobody, or a director describing commercial achievement, that is the first thing to establish rather than the last.
5. Check whether the evidence still exists. Our record-keeping checklist for engineering projects and our nine-part framework for R&D project records set out what a defensible file looks like. A period where the file is thin is not automatically a period where the claim was wrong, but it is where you should look first.
6. Then decide. The options run from doing nothing on a sound review, through amending a return still within time, to a disclosure to HMRC. Which one applies depends on the facts and on the behaviour behind the inaccuracy, and it is a decision to take with advice.

If a letter has already arrived, the sequence is different and the priority is the response rather than the review. Our blog on R&D enquiries and how to respond covers that ground.

How Lexmore approaches this

We review historic claims project by project against the statutory definition, using the same eligibility work we would apply to a new claim. That means reading the narratives that were actually submitted, identifying the competent professional behind each project, and testing whether the advance and the uncertainty can still be evidenced.

Most reviews of this kind end with a company knowing its position and doing nothing further. Where they do not, the route depends on the facts and on the behaviour behind the inaccuracy, and we would set out the options rather than push a particular one. What we would not do is tell a company its historic claims are fine because nobody has asked about them.

Lexmore’s View

A decade of accepted claims is a decade of not being selected. For much of that decade the odds of selection were low by design, because HMRC did not have the people to do otherwise, and its own random sampling later put error and fraud in the SME scheme at around a quarter of the relief claimed. Those two facts sit together and they explain each other.

None of that means your claims were wrong. It means HMRC’s silence was never evidence that the underlying claim was correct. The company carries the responsibility whoever prepared the return, the discovery window is longer than the enquiry window by some years, and the gap between an unprompted disclosure and a prompted one is worth real money. Knowing where you stand. Deciding on the facts. Choosing the timing yourself.

References

• HMRC: Approach to Research and Development tax reliefs 2023 to 2024
• HMRC’s approach to Research and Development tax reliefs (2023)
• HMRC annual report and accounts 2025 to 2026: Our accountability
• CIRD80525: Examining a claim, service aims
• EM1510: Opening the enquiry, CTSA time limits
• CH56200: Assessing time limits, Corporation Tax
• CH82470: Maximum and minimum penalties for each type of behaviour
• CH81125: Reliance on another person
• CH81165: Inaccuracy attributable to another person
• CH880100: Dishonest tax agents, overview
• GfC3 Part 2: Expectations of claimants
• HMRC interest rates for late and early payments