The line, in HMRC’s own examples
HMRC publishes no engineering-specific R&D guidance page — unlike software and pharmaceuticals, which have their own CIRD chapters — so this page runs on the general DSIT definition and GfC3’s worked examples.
Two GfC3 examples bracket the sector. At the non-qualifying end sits an experienced engineer refining working practices: real value, but no claim, because the improvements were not developed as part of a specific project seeking an advance in technology. At the qualifying end sits the refinery-catalyst case in chemical engineering. It qualifies from the point the technological uncertainties are identified, through the systematic work to resolve them. Alongside them, the warehouse-robotics example splits a single programme. The engineers’ qualifying development sits on one side, and the marketing department’s research, expressly not qualifying, on the other.
There is an uncomfortable truth here: the better your engineers, the higher the bar, because readily deducible is measured against competent professionals exactly like them. So engineering claims should be narrow and deep. Claim the sub-projects where your own experts genuinely did not know the route, not a broad coat of “innovation” across delivery work.
Design, development, delivery
Most engineering businesses run all three, and the claim lives almost entirely in the middle:
- Design within established methods, however demanding the specification, is the profession’s baseline. Codes, published data and standard modelling applied well do not qualify.
- Development against genuine unknowns is the qualifying core. Methods run out, and the team hypothesises, models, prototypes, tests and iterates. The window opens when the uncertainty is identified and closes at resolution, the same day-counting discipline HMRC applies elsewhere.
- Delivery and validation of the settled solution sits outside, save for genuine qualifying indirect activities. Manufacture support, routine testing and documentation are the usual examples.
Sector pitfalls
- Difficulty inflation. Tolerance-chasing and integration pain presented as uncertainty; the test asks what the field could do, not how hard the week was.
- Bespoke-equals-R&D. One-off machines and rigs built from known engineering are unique products, not advances, the same trap as construction’s bespoke sites.
- The client-money blind spot. Development contemplated in a client’s contract, claimed as your own. Since 2024 that is an entitlement error, not a nuance. The rules say whose claim it is.
- Unwritten judgement. Engineering happens in models and meetings; a claim without the failed-iteration record meets the evidence bar as assertion. The chief engineer’s written reasoning, kept, is the cheapest insurance in the sector.
Frequently asked questions
Our engineers solve difficult problems daily. Why is that not all R&D?
Because the test measures the field, not the difficulty. HMRC's example of an engineer refining working practices fails precisely because it was not a project seeking an advance; expertise applied well is the baseline, and R&D starts where that baseline demonstrably runs out.
Does designing to extreme specifications qualify?
Not by the extremity alone. Designing within established methods to a hard spec is competent engineering. It qualifies when meeting the spec defeats readily available knowledge, as in HMRC's refinery-catalyst example, which qualifies from the point the uncertainties are identified.
Our development work is paid for by clients. Whose claim is it?
The 2024 entitlement rule decides: generally whoever decided to undertake the R&D, read from the contract and circumstances. Client-commissioned development contemplated by the contract sits with the client; your own initiated R&D inside a delivery contract can be yours. Engineering consultancies should treat this as a contract-drafting issue now.
Do simulations and testing qualify?
Inside the uncertainty window, modelling, prototyping and instrumented testing are classic qualifying activities, and devising new testing methods can qualify in its own right under the guidelines' indirect-activity provisions. Routine validation of a settled design is outside.