R&D tax relief was always meant to reward genuine innovation — companies doing the hard, uncertain technical work that advances what is scientifically or technologically possible. For a long time, the claims environment was permissive enough that loose standards slipped through. That era is over. HMRC has significantly tightened its enquiry activity, and the statistics reflect it: SME scheme relief fell by 29% in 2023/24, and more than three-quarters of the claims that were examined were adjusted or rejected outright.

The question of how UK tech startups can strengthen their R&D tax relief claims under current HMRC rules is one we get asked regularly — and the honest answer is that the underlying rules have not become dramatically more complex. What has changed is how hard HMRC looks, and how little patience it has for claims that rely on assertion rather than evidence. The startups that are getting this right are not doing anything exotic. They are being methodical from the outset.

Here is how we think about building a robust R&D claim in the current environment.

The qualifying threshold for R&D relief has always rested on two tests: the project must seek an advance in science or technology, and it must involve genuine scientific or technological uncertainty. That second limb — uncertainty — is where a significant number of startup claims run into trouble.

HMRC's definition of technological uncertainty is precise: it exists when a competent professional in the relevant field could not determine in advance whether the outcome was achievable or how to achieve it. Adapting existing technology, improving a known process, or building something difficult-but-well-understood does not clear that bar. HMRC caseworkers increasingly test this through detailed technical questions, and a claim that cannot answer them credibly will not survive.

From 1 April 2023, mathematical advances can be treated as science for R&D purposes, which is useful for startups working on algorithms, modelling, or data-intensive products. But the advance still needs to be demonstrable, not assumed.

The practical takeaway: before you prepare a claim, ask honestly whether a senior engineer or scientist in your space would, at the time the project started, have been uncertain whether what you were attempting was technically achievable. If the honest answer is no — it was hard, time-consuming, or expensive, but not genuinely uncertain — you may not have an R&D project. Building a claim on uncertain ground is not just risky; it is increasingly likely to cost you more than the relief you recover.

The single biggest practical gap we see in startup R&D claims is the quality and timing of documentation. A claim written retrospectively — based on a one-page narrative assembled six months after the accounting period closed — is almost impossible to defend at enquiry. HMRC is experienced at identifying post-rationalisation, and it flags vague language, sweeping assertions, and narrative gaps accordingly.

Strong documentation captures the uncertainty and the iterative process of resolving it in real time. That means:

• Technical specifications or hypothesis documents written before or during the project, not after
• Contemporaneous notes of failed approaches, pivots, and the reasons for them
• Sprint or project management records that show the experimental cycle — hypothesis, test, result, iteration
• Internal communications (Slack, email threads, engineering tickets) that demonstrate the team was genuinely working through unknown territory

This is not about creating paperwork for HMRC's benefit. It is about running the kind of engineering discipline that good technical teams run anyway — and ensuring that record is preserved and accessible when the claim is prepared. Version control systems, Jira boards, and engineering wikis already contain most of what you need if you know to look for it.

The R&D claim narrative itself should tell the story of the uncertainty and how the team resolved it. HMRC is sceptical of claims that describe what was built without explaining why it was technically difficult. The two questions the narrative needs to answer are: what did you not know at the start, and what competent professional would have shared that uncertainty?

Staffing costs are typically the largest qualifying category in a tech startup R&D claim — and they are also the category most likely to be challenged or adjusted at enquiry. Optimistic staff apportionment is one of the most consistently cited reasons for claims being reduced.

The issue is straightforward: only the time actually spent on qualifying R&D activity can be included. A developer who spent 40% of their time on R&D and 60% on building, maintaining, or supporting a commercial product cannot have 80% of their salary included in the claim. HMRC expects to see a credible, defensible rationale for the percentages applied — ideally backed by timesheets, project logs, or a structured methodology for calculating the split.

For early-stage startups where roles are fluid and people wear many hats, this is genuinely complicated. A CTO who contributes to qualifying R&D, product management, client delivery, and recruiting in a single month needs a considered approach to apportionment — not a round number that bears no relation to how their time was actually spent.

Our view is that building a time-tracking process specifically for R&D purposes is worth the overhead, even if the firm does not track time for any other reason. It does not need to be granular to the hour. A monthly or project-level record of what qualifying work was done, by whom, and for how long is enough to put the claim on defensible ground.

Subcontractor costs and externally provided workers are also qualifying categories, but the eligibility rules here have tightened further under the merged scheme, particularly for overseas subcontractors. If you are using a development agency or offshore team for any part of the qualifying work, check the current rules carefully before including those costs.

From 1 April 2024, the previous separate SME and RDEC schemes were merged into a single R&D expenditure credit (RDEC) regime for most companies, with a higher rate of relief available to loss-making, R&D-intensive SMEs under what is now called the Enhanced R&D Intensive Support (ERIS) scheme. This is a material structural change, and it affects how the relief interacts with your tax position depending on whether your startup is profitable, loss-making, or somewhere in between.

The merged scheme is credit-based rather than deduction-based, which changes the cashflow mechanics of the relief. Under the old SME scheme, a loss-making startup could often receive a payable cash credit relatively quickly. Under the merged RDEC regime, the credit is applied against Corporation Tax first, with any excess potentially payable — but the sequencing matters and the net benefit depends on your company's specific tax position.

ERIS is worth examining if your startup is genuinely R&D-intensive — broadly, if qualifying R&D expenditure represents at least 30% of total expenditure. The enhanced credit rate here is more favourable, but the eligibility conditions need to be met properly and documented. If you are close to the threshold, the apportionment of costs between R&D and non-R&D activity becomes even more significant.

The scheme changes are one reason we are cautious about startups relying on claim templates or advisers whose knowledge predates 2024. The structural mechanics of how relief is calculated and paid have changed, and a claim prepared under old assumptions may produce incorrect figures even if the underlying R&D activity is entirely legitimate.

HMRC maintains an informal list of what it regards as high-risk R&D claim agents — advisers whose claims are disproportionately likely to be inaccurate or inflated. Being associated with an adviser of that kind is itself a red flag that increases the likelihood of an enquiry, irrespective of the quality of your underlying claim.

The R&D advisory market grew rapidly during the period of looser scrutiny, and some of that growth was driven by firms charging success fees to submit claims of questionable quality. Several of those firms are no longer operating following HMRC enforcement action, but the market still contains advisers whose approach does not reflect the current compliance environment.

When choosing an adviser for an R&D claim, we would look for:

• A firm that asks detailed technical questions before committing to a claim — not one that assures you the claim will go through before understanding your business
• Transparent fee structures — success-only fees on R&D claims are not inherently wrong, but they create an incentive to inflate claims that you should be aware of
• A track record that includes handling HMRC enquiries, not just submitting unchallenged claims
• Professional indemnity and a clear position on what happens if the claim is challenged

R&D relief is a legitimate and valuable incentive for genuine innovation. The startups that are building strong claims in the current environment are not gaming the system — they are doing the technical work, keeping contemporaneous records, and working with advisers who understand the current rules. That combination holds up under scrutiny.