Short answer
Defence R&D qualifies on exactly the same test as any other sector — there is no defence exemption, and no defence concession either. What is different is evidential. HMRC requires a written description of each project’s advance and uncertainty, and a great deal of defence work is classified, export-controlled or commercially restricted. HMRC has published nothing on how to handle that, so the burden sits entirely with the claimant to describe the technology without describing the application. It is usually possible. It has to be done deliberately.
Applies to
- Schemes
- All periods · Merged scheme · ERIS · Legacy SME · Legacy RDEC
- Periods
- 1 April 2023 onwards
- Claimants
- All
- Sectors
- Defence
This note covers defence primes, subsystem and component suppliers, dual-use technology businesses, and companies working to MOD, DASA or allied programmes.
Where the R&D usually is
The qualifying work is where a capability requirement could not be met by anything available, and it was not clear at the outset whether or how it could be met.
In practice that means systems and subsystems developed against requirements no existing product satisfies — detection, discrimination, endurance, survivability or resilience thresholds that nothing on the market reaches. It means materials, structures and manufacturing methods developed for environments existing ones cannot survive. It means software, autonomy, signal and data processing where the behaviour of the integrated system is not deducible from its parts. And it means hardening and assurance work where meeting a resilience requirement needs a genuine technical advance rather than the application of a known technique.
Two things carry across from adjacent sectors. The advance need not be visible in the delivered capability — a process or manufacturing advance counts as much as a product one. And a project that fails still qualifies for the period in which the work was done, because the test is the advance sought, not the advance achieved.
Claiming work you are not allowed to describe
This is the sector’s defining problem, and it has no published solution.
HMRC’s additional information form requires, for each project claimed, a description of the advance sought, the uncertainties faced and how they were addressed. That obligation does not bend for classification, and you cannot satisfy it by cross-referring to a report you are not able to send. See the additional information form.
The good news is that the statutory test helps. The test asks about capability in a field of science or technology — not about the platform, the customer, the operational requirement or the performance figures achieved. So an advance can almost always be stated at the technological level while the application stays out of it. A description that says what class of problem was unsolved, why existing approaches failed, what was attempted and what remained unknown can be complete on its own terms and disclose nothing restricted.
What does not work is vagueness. “We developed advanced capability for a classified customer” is not a description; it is a refusal to give one, and it reads to a caseworker exactly like a claim with nothing behind it. The distinction that matters is between withholding the application, which is legitimate and usually invisible in a well-written narrative, and withholding the technology, which leaves nothing for the claim to stand on.
Two practical points follow. Get the narrative cleared internally — by security and export control, not only by the engineers — before it goes to HMRC, and build that step into the claim timetable rather than discovering it at the deadline. And keep the underlying technical record in a form your own people can access under enquiry conditions, because if HMRC asks for supporting evidence, “we have it but cannot show you” is a materially worse position than a narrative that never needed it.
Where a genuine conflict remains — where the technology itself cannot be described without disclosing something restricted — there is no published route through it, and the honest position is that this is unresolved rather than solved. Take advice early, and expect to discuss it with HMRC rather than to find a rule.
Where development ends
Defence programmes run for years through demonstration, qualification and into service, and the claim is a subset of each period.
R&D ends when the uncertainty is resolved, or work on it stops. Qualification against a defence standard, environmental and safety testing of a settled design, trials to demonstrate a known capability to a customer, and the documentation supporting acceptance are all evidencing rather than resolving — the same boundary that governs aerospace, and for the same reason. Where trials expose a shortfall, and engineers develop a fix, that development is fresh R&D.
Technology readiness levels are a useful internal shorthand for where a programme sits, and low levels usually coincide with qualifying work while high ones usually do not. But readiness level is not HMRC’s test and no claim should be built on it. The question is always whether a competent professional could have said at the outset how to achieve the thing.
Primes, subcontracts and who claims
Nearly every defence business sits in a chain, and entitlement decides the claim.
For accounting periods beginning on or after 1 April 2024, the merged scheme asks who intended and contemplated the R&D. Where a prime issues a requirement and leaves you to work out how to meet it, the development is usually yours; where you are engaged to carry out a development programme the customer has defined and directed, it usually is not. The full test is in contracted-out R&D.
Defence adds a wrinkle worth naming: contracts are often more prescriptive than in civilian sectors, and a detailed requirement specification is easily mistaken for a commissioned development. It is not the same thing. A requirement states what the capability must do; a commissioned development tells you what to build and how. The question is which one you were given.
Costs that behave differently
Most cost rules apply as they do anywhere — see what costs qualify. Three points are particular to this sector.
Security clearance and vetting costs are not R&D. They are a condition of doing the work, not an activity that resolves a technological uncertainty, and the guidelines’ list of qualifying indirect activities is closed.
Dual-use programmes need the boundary drawn once, properly. Where a technology has civil and defence variants developed from common work, the qualifying activity is the common technical development — and the variant-specific integration is usually not.
Overseas work packages need care. For accounting periods beginning on or after 1 April 2024, the restriction reaches payments to contractors for R&D undertaken abroad and externally provided workers outside UK PAYE, but not your own employees. Where it applies, expenditure can still qualify if conditions necessary for the work are not present in the UK. Cost and the availability of workers are expressly excluded and are the only exclusions — so “the programme required an allied facility” may be arguable where “it was cheaper” is not.
Worked example
Illustrative. A subsystem supplier develops against a capability requirement, then qualifies and delivers.
| Workstream | Spend | In the claim? | Why |
|---|---|---|---|
| Development against a detection threshold nothing available meets | £380,000 | Yes | No deducible approach; resolved by experiment |
| Rig and environmental testing feeding back into the design | £110,000 | Yes | Testing to resolve the uncertainty |
| Redesign after trials showed the threshold was not met | £130,000 | Yes | A further uncertainty the trials exposed |
| Qualification against the applicable defence standard | £165,000 | No | Demonstrating a settled design complies |
| Security clearance and vetting for the programme team | £40,000 | No | A condition of the work, not the work |
| Production of delivered units | £220,000 | No | Manufacturing |
Of £1,045,000, £620,000 sits inside the claim. The £130,000 redesign is the row most often lost because programmes book it to the trials phase.
Where claims go wrong
- Writing a narrative so guarded it says nothing. The commonest defence failure. A description that withholds the technology as well as the application gives a caseworker no basis to accept the claim and looks indistinguishable from a claim with nothing behind it.
- Leaving the security review to the end. The narrative needs clearance from security and export control as well as engineering. Built into the timetable, it costs days; left to the deadline, it costs the claim notification or the filing window.
- Cross-referring to a document you cannot send. The form has to stand alone. It always did, and here there is no alternative.
- Claiming the qualification programme. Testing to a defence standard on a frozen design is evidencing compliance, not resolving uncertainty — and the redesign that testing sometimes forces is the part that qualifies.
- Reading a requirement specification as a commissioned development. Defence contracts are prescriptive by nature. Prescriptive about what is not the same as directive about how, and the difference decides who claims.
- Assuming records kept under security constraint will be available later. If an enquiry opens two years on, the claim is defended from what can actually be retrieved and shown — see record-keeping requirements.
John Mapother
Founder & Director
Last reviewed 14 September 2026