R&D tax credits UK claims often include sizeable software licences, cloud compute and data spend. These costs are frequently mixed-use, so the claim succeeds or fails on whether your apportionment is reasonable, repeatable and evidence-backed. HMRC’s own manuals explicitly anticipate pragmatic apportionments for software used partly in R&D. [1]
Table of Contents
What qualifies, at a high level
Software: revenue expenditure on software employed in R&D can qualify, including use in qualifying indirect activities. Where software is only partly employed in direct R&D, an appropriate apportionment should be made. HMRC notes that a pragmatic basis, such as staff numbers, can be suitable and that detailed enquiries are not envisaged if the basis is reasonable. [1]
Data licences and cloud computing services: these costs can qualify when employed in activities that directly contribute to resolving scientific or technological uncertainty. Where the spend relates to qualifying indirect activities rather than direct R&D, it does not qualify. [2]
Apportionment principles to follow
- Attribute to activities, then to projects. Start with what the service was used for (direct R&D, qualifying indirect activity, or non-R&D).
- Prefer objective metrics over judgement calls. Usage data beats estimates, even if the metric is imperfect.
- Document why the method is reasonable for your facts. HMRC signals a preference for pragmatism where it produces a fair result. [1]
- Make it reproducible. A good apportionment can be rerun by someone else using the same data sources.
Five allocation models that usually work
- Environment separation: Split cloud accounts or resource groups into R&D and production environments, then claim 100% of the R&D environment costs and 0% of production. Use tagging policies and monthly exports as evidence.
- Usage-based metering: Allocate compute, storage and data transfer using provider usage reports (for example, by project tag, workspace, or cluster).
- Staff-ratio apportionment: For shared licences, allocate cost by the ratio of R&D staff to total users. HMRC explicitly cites staff-number apportionment as pragmatic for mixed-use software. [1]
- Time-window allocation: Where R&D activity is concentrated in sprints, allocate costs to those time windows using deployment logs, Jira exports or build pipeline records.
- Output-based sampling: Use a documented sample of services or datasets, trace them to projects, and extrapolate with a clear rationale. Keep the sample selection logic and the underlying extracts.
What not to do
- Claim 100% of a general cloud bill because ‘the company is an R&D business’.
- Use a single flat percentage across all software and cloud spend without a metric or evidence trail.
- Treat data acquisition as qualifying when it is not linked to direct R&D activity or when it supports non-R&D analytics. [2]
AIF-ready evidence pack for digital costs
The AIF requires project narratives and project-level cost totals. That means your apportionment has to land in a project register, not just a spreadsheet. [3] A practical evidence pack for software, cloud and data typically includes:
- Supplier contract or invoice list showing what was bought and the billing period.
- Usage exports (cloud cost and usage reports, tagging extracts, licence user lists).
- Allocation logic document (one page): metric, formula, inclusions and exclusions, and how it links to the project register.
- Management sign-off by the named internal R&D contact responsible for the claim (the AIF asks for this contact role). [3]
Where FI Group by EPSA fits
FI Group by EPSA supports businesses in building defensible methodologies for qualifying expenditure, including pragmatic apportionments for mixed-use software and cloud services, and in packaging those methods into AIF-ready evidence. For a general overview of the UK relief, see this R&D tax credits UK overview.
FAQs
Can we claim cloud costs that support indirect activities?
The HMRC manual for data and cloud costs states that where such costs are attributable to qualifying indirect activities rather than direct R&D, the expenditure does not qualify. [2]
Is staff-number apportionment acceptable for shared software licences?
HMRC’s software guidance describes staff-number apportionment as a pragmatic approach where software is used by both R&D and non-R&D staff. [1]
Do we need tagging to claim cloud spend?
Not always, but tagging or other objective usage metrics make apportionment easier to evidence and reproduce. If you cannot tag, document an alternative metric and keep the underlying extracts.
What is the minimum evidence we should retain?
Invoice list, usage exports, allocation logic, and a clear link to project-level totals used in the AIF. [3]
How do we handle datasets used by both R&D and production?
Separate where possible. If not, allocate by measured usage (for example, compute hours consumed in R&D environments) and document the method.
