Short answer
An advance is an increase in the overall knowledge or capability in a field of science or technology — what the field as a whole knows or can do. It is measured against knowledge that is publicly available, or readily deducible from publicly available knowledge, by a competent professional working in that field. It is not measured against what your company knew, what your competitors sell, or what is new to your industry.
Applies to
- Schemes
- Merged scheme · ERIS · Legacy SME · Legacy RDEC · All periods
- Periods
- 1 April 2023 onwards
- Claimants
- All
The advance is in the field, not in your company
The guidelines are explicit that an advance means an advance in overall knowledge or capability in a field of science or technology, “not a company’s own state of knowledge or capability alone”. Everything else follows from that sentence.
It means a company that buys in a capability its competitors already have is not doing R&D, however transformative that is for the business. It means the first firm in a sector to adopt a technique used routinely in an adjacent sector is usually not doing R&D either. And it means the question “was this hard for us?” is the wrong question. The right question is “could a competent professional in this field already have done this, or worked out how to?”
The four shapes an advance takes
The guidelines give four examples of a project that will be R&D. A project that seeks to:
- extend overall knowledge or capability in a field of science or technology;
- create a process, material, device, product or service which incorporates or represents an increase in overall knowledge or capability;
- make an appreciable improvement to an existing process, material, device, product or service through scientific or technological change; or
- use science or technology to duplicate the effect of an existing process, material, device, product or service in a new or appreciably improved way.
The fourth is the one people miss. Reproducing something that already exists can be R&D where you are doing it in a genuinely different way and the details of how it was originally done are not publicly available. Reverse-engineering an outcome from first principles is a real R&D pattern; buying the product and copying it is not.
Appreciable improvement, and what is not one
Most claims rest on the third shape, so the meaning of “appreciable improvement” carries a lot of weight. It means changing or adapting the scientific or technological characteristics of something to the point where it is genuinely and non-trivially better — not a minor or routine upgrade.
Two things are expressly excluded:
- Deployment in a new context. Taking existing science or technology and using it somewhere new, in a different trade for instance, with only minor or routine changes, is not an appreciable improvement.
- Catching up. Something is not appreciably improved if the work simply brings your company into line with what the field can already do — even though it may be completely new to you and to your trade.
What counts as appreciable varies between fields. In a mature, heavily optimised field a half-percent efficiency gain may be a real advance; in a young field it would not be. That judgement belongs to a competent professional working in the field, which is why the identity of that person matters as much as the description of the work.
You do not have to succeed, and you do not have to be first
Two points that consistently surprise people.
Failure still counts. Even if the advance sought is not achieved, or not fully achieved, R&D still took place. The relief is for the attempt to resolve uncertainty, not for the result. In practice an abandoned project with good contemporaneous records is often easier to defend than a successful one, because the uncertainty is evident on the face of the file.
Someone else may have got there first. If an advance has already been made or attempted but the details are not publicly available — because they are a trade secret, for example — work to achieve that advance can still be an advance in science or technology. What matters is what is in the public domain, not what exists somewhere in the world.
The corollary is uncomfortable and is worth stating plainly: once the advance does enter the public domain, the same work stops qualifying. A project that qualified in its first year may not qualify in its third.
Routine work is not an advance
Routine analysis, copying or adaptation of an existing process, material, device, product or service does not advance overall knowledge or capability. This is the line that most rejected claims fall the wrong side of.
The distinction is not about effort. Assembling a large system from well-understood components is often a substantial engineering job, and it is not R&D if a competent professional would have known the approach would work. The tribunals have taken a narrow view here: using existing components, libraries or frameworks does not by itself disqualify a project, but the advance has to be somewhere other than in the assembly.
Position for accounting periods beginning before 1 April 2023
The earlier guidelines, issued in 2004 and updated in 2010, define an advance in the same terms, and the tribunal decisions on appreciable improvement decided under them apply to the current guidelines without adjustment.
The difference is the scope of “science”. For those earlier periods, a mathematical advance was within scope only where it advanced the representation of the nature and behaviour of the physical and material universe. Work whose advance is purely mathematical falls outside the definition for accounting periods beginning before 1 April 2023 and inside it for periods beginning on or after that date.
Worked example
Illustrative. Two companies in the same year, each spending around £400,000 on development.
| Company A | Company B | |
|---|---|---|
| The work | Replacing a manual scheduling process with an off-the-shelf optimisation package, configured to the company’s operations | Developing a scheduling algorithm for a constraint structure that published optimisation methods handle only by approximation, where accuracy matters |
| New to the company? | Yes, entirely | Yes |
| New to the field? | No — the package is commercially available and the configuration is documented | Yes — no published method addresses this constraint structure directly |
| Competent professional’s view | Would have known this would work, and how | Could not say in advance whether an exact method was achievable |
| Advance? | No | Yes |
Company A has made a large, valuable change to its business and has no R&D claim. Company B has a claim whether or not the algorithm ends up outperforming the approximation, because the uncertainty was real when the work started.
Where claims go wrong
- Describing a product instead of a capability. “We built a platform that does X” states an outcome. The advance is what the field could not do before, expressed as knowledge or capability. If the description would read the same in a sales brochure, it is not an advance statement.
- Benchmarking against competitors. “No one else in our sector offers this” is a commercial claim, not a technological one, and it is often true of work that any competent professional could have done. The benchmark is the field’s knowledge, not the market’s product range.
- Claiming the integration. Connecting established systems to each other is the most frequently over-claimed pattern we see. It can involve a genuine advance, but the advance has to be identified specifically — not asserted from the fact that the integration was complicated.
- Letting the advance drift. Where the description given to HMRC at enquiry does not match the one given at the outset, the claim is in serious trouble. Tribunals have treated inconsistent accounts of what the project was as evidence that the later account was constructed.
- Claiming the same advance every year. Once the knowledge is public — including where the company itself has published or shipped it — the advance has been made. Rolling a claim forward on the same narrative for three years is a reliable way to attract an enquiry into all three.
John Mapother
Founder & Director
Last reviewed 31 August 2026