R&D tax relief: does your startup actually qualify?

R&D tax relief has a branding problem. Half of founders assume it is only for people in lab coats; the other half were told by a cold-caller that their café's new loyalty card qualifies. Both are wrong, and the second group is the reason HMRC now opens enquiries into claims that would have sailed through five years ago.

The rules also moved under everyone's feet. For accounting periods beginning on or after 1 April 2024, the old SME scheme and RDEC were replaced by a single merged R&D expenditure credit at 20%, sitting alongside a more generous route called Enhanced R&D Intensive Support (ERIS) for loss-making companies that spend most of their money on R&D. This is the founder's version: whether your work qualifies, what a claim is worth in pounds, and the two deadlines that kill more claims than HMRC enquiries ever do.

The actual test — and the one question that settles it

A project qualifies if it seeks an advance in science or technology by resolving scientific or technological uncertainty: uncertainty that a competent professional working in the field could not readily deduce their way out of. Two things follow from that wording, and both catch founders out.

First, the advance has to be an advance in the field, not an advance for your company. Solving a problem your team had never met before, but which any senior engineer in that discipline could have sketched on a napkin, is not an advance. Second, commercial novelty is irrelevant. "The first meal-kit service aimed at climbers" is a business idea. HMRC is asking a technical question, not a market one.

For software startups, where most of the arguing happens:

• Usually qualifies: algorithms with no published solution; making a system perform at a scale where the standard architecture demonstrably falls over; getting systems to interoperate where the integration itself required experiment, failure and redesign; machine learning work where whether the approach would function at all was genuinely unknown at the outset.
• Usually does not: building a standard web or mobile app with mature frameworks; interface design and styling; configuring off-the-shelf software; API integrations that worked as documented; data migration; and testing and bug-fixing that was hard work rather than uncertain work.

The honest tell is a question for your technical lead rather than your accountant: on this project, was there a point where nobody knew whether it could be done at all, and you had to run experiments to find out? If the answer is yes, and someone wrote down what was tried and what failed, that is your claim. If the answer is "it was just a lot of work", it is not, and no amount of narrative writing will make it one.

What a claim is actually worth

Two routes, two very different numbers.

The merged scheme — 20%, but taxable

The merged R&D expenditure credit is 20% of qualifying expenditure. The credit is itself taxable, so what you keep depends on your corporation tax position. A company paying the 25% main rate nets 15p per £1 of qualifying spend. A company on the 19% small profits rate, or loss-making — where the credit is restricted using the small profits rate — nets 16.2p.

ERIS — the loss-making, R&D-intensive route

If your company is loss-making and qualifying R&D is at least 30% of total expenditure, you can claim ERIS instead. You deduct an extra 86% of qualifying costs (186% in total), surrender the resulting loss, and take a payable credit at 14.5%. That comes to 26.97p per £1 of qualifying spend, which is why the intensity calculation is worth doing properly before anything is filed. There is also a year of grace: a company that met the intensity condition in one period can still use ERIS in the next even if it narrowly fails the test.

Qualifying costs are mostly people: the share of each person's salary, employer National Insurance and pension that went into the R&D itself. Then externally provided workers and subcontractors at restricted rates, consumables genuinely used up, software licences, and cloud computing and data costs. Rent, marketing, patent filing and the founders' time spent fundraising are not qualifying, however central they felt at the time.

The PAYE and NIC cap

Both routes cap the payable credit at £20,000 plus 300% of the company's relevant PAYE and National Insurance liabilities for the period. It exists to stop companies with no real UK employment footprint extracting cash, and it bites hardest on exactly the startups that need the money: those whose engineers all invoice through their own limited companies, or whose founders take dividends rather than salary. If your PAYE and NIC bill is small and your R&D spend is large, model the cap before you write the credit into a cashflow forecast. How you pay yourself is not a neutral decision here — our guide to paying yourself as a founder covers the trade-off from the other direction.

The two deadlines that lose the money

Neither has anything to do with whether your work qualifies, which is what makes missing them so galling.

1. Claim notification — six months after the end of the period of account. For accounting periods beginning on or after 1 April 2023, a company claiming for the first time, or one that has not claimed in the previous three years, must submit a claim notification form. The window opens on the first day of the period of account and closes six months after it ends. For a year ended 31 March 2027, that is 30 September 2027. Miss it and the year is gone, however strong the science was.
2. Additional information form — before, or on the same day as, the CT600. Mandatory for claims submitted on or after 8 August 2023. Send the tax return first and HMRC writes to confirm it has removed the R&D claim from your return. The form asks for your UTR, PAYE reference, VAT number and SIC code, the accounting period dates, the senior internal contact responsible for the R&D, details of every agent involved, a breakdown of qualifying costs, and written project descriptions.

How many projects you must describe depends on how many you have: with one to three, describe all of them; with four to ten, describe at least three covering 50% or more of the qualifying spend; with more than ten, describe up to ten. The claim itself must be made within two years of the end of the accounting period, so a late notification cannot be rescued by a fast claim.

A word on the claims industry

The "no win, no fee, everyone qualifies" boutiques exist because for a decade the scheme was policed lightly. It is not any more. HMRC's compliance teams recognise the templates, and an enquiry lands on your company, not the adviser who wrote the narrative — you carry the repayment, the interest and any penalty. A claim built on a real conversation with your engineers, with contemporaneous notes behind it, survives scrutiny. A claim assembled from a questionnaire and a thesaurus does not.

What to do this week

1. Put both deadlines in the calendar now. Take your accounting period end, add six months for the notification date, add two years for the claim deadline. Set reminders three months before each.
2. Open a project log. One page per R&D project: what you were trying to achieve, why the answer was not already known, what you tried, what failed, what you learned. Ten minutes a fortnight now is worth days of reconstruction later.
3. Tag engineer time to projects. Whatever tracker you already use is fine. Apportioning salaries a year later always produces a smaller, weaker claim than recording it as you go.
4. Run the intensity number. Qualifying R&D divided by total expenditure. If it is anywhere near 30%, the ERIS-versus-merged-scheme difference is worth real money and deserves a proper calculation.
5. Check the cap arithmetic. £20,000 plus three times your PAYE and NIC. If that is lower than your expected credit, adjust the cashflow forecast before it disappoints you.

The bottom line

R&D relief is one of the few pieces of the UK tax system genuinely designed to fund early-stage technical companies, and for a loss-making startup at 26.97p in the pound it can be the cheapest money you ever raise — cheaper than the equity routes set out in our funding guide and with none of the dilution described in our post on SEIS and EIS for founders. But it is unforgiving on process. We assess qualifying work honestly, including telling you when it does not qualify, build the narrative with your technical lead, and file it alongside your corporation tax return — the wider picture is in our guide to corporation tax for startups. If you are spending real money on hard technical problems, get started and let us look at it properly.