What Is an R&D Project? Setting the Start and End Points of Your Claim

Every R&D claim is built from projects, so before a company can claim r&d tax relief it has to answer what is an R&D project in its own case and map the R&D project stages, from the first work on the uncertainty to its resolution. The R&D projects examples in the government's guidelines show where each boundary falls.

A first claim carries more risk than a repeat one. HMRC audited a random sample of 500 SME claims received in 2020 to 2021 and found that a quarter (25%) were fully disallowed because no qualifying R&D activity took place, that 19% had overclaimed the relief, and that 38% of the value of first-time claims was non-compliant against 22% for companies that had claimed before (HMRC, "HMRC's approach to Research and Development tax reliefs", 17 July 2023). A misdrawn project boundary produces both kinds of error, and it sits alongside the other mistakes that get R&D tax relief claims rejected.

What Is an R&D Project for Tax Relief?

A project "consists of a number of activities conducted to a method or plan in order to achieve an advance in science or technology" (DSIT Guidelines 2023, para 19, reproduced in HMRC's manual at CIRD81910). The project is defined by the uncertainty the work tries to resolve, whatever budget code or launch date the business attaches to it.

The same paragraph says the project "should encompass all the activities which collectively serve to resolve the scientific or technological uncertainty associated with achieving the advance". It may sit inside a larger commercial project, "but that does not make the parts of the commercial project that do not address scientific or technological uncertainty into R&D". The commercial go-ahead, the business case, set-up and marketing stay outside the claim even when they share a budget with the R&D.

The boundary only matters once the work passes the qualifying test, which asks whether a competent professional in the field could readily resolve the uncertainty (CIRD81910, paras 13 and 14). The guide to everyday problem-solving versus qualifying R&D covers that test.

R&D Project Stages: Where the Work Starts and Stops

The guidelines set four boundary rules (CIRD81910, paras 33 to 35 and 38):

1. Start: R&D begins when work to resolve the scientific or technological uncertainty starts (para 33).
2. End: R&D ends when that uncertainty is resolved or work to resolve it ceases (para 33), when the knowledge is codified in a form a competent professional can use, or when a prototype or pilot plant with all the functional characteristics of the final product is produced (para 34).
3. Failure: a project that does not succeed still counts if it set out to achieve an advance (para 38).
4. Restart: a new uncertainty that emerges after the end can begin new R&D, which the guidelines distinguish from routine fault fixing (para 35).

On a commercial timeline, the R&D project stages form a window: the idea and the go-ahead come before it, technical planning, design, build and testing sit inside it, and production, rollout and support follow it.

Where the Project Starts: First Work on the Uncertainty

R&D "begins when work to resolve the scientific or technological uncertainty starts" (CIRD81910, para 33). The same paragraph excludes "work to identify the requirements for the process, material, device, product or service, where no scientific or technological questions are at issue".

Technical scoping is inside: defining technical objectives, assessing feasibility and identifying the particular uncertainties are R&D (para 36). Market research and a project's financial, marketing and legal aspects are not (para 37). Take the start date from records made at the time, such as meeting notes, tickets and lab books.

Where the Project Stops: Resolution or Abandonment

R&D ends "when that uncertainty is resolved or work to resolve it ceases" (CIRD81910, para 33). For a prototype, that point comes once the changes the test findings call for have been made and further testing has been satisfactorily completed (para 39).

Deciding when to stop a project is a commercial call, and the R&D ends on the day work on the uncertainty stops, so a project that "never really ended" still has an end date. Abandonment does not void the claim, because what counts is the intention to achieve an advance (para 38).

Scoping, Testing and Rollout: On Either Side of the Boundary

Activity	Inside or outside	Source (CIRD81910)
Market research, business case, commercial go-ahead	Outside	paras 28(a) and 37
Requirements gathering with no technical question at issue	Outside	para 33
Technical planning: objectives, feasibility, identifying uncertainties	Inside	para 36
Feasibility studies informing the direction of a specific R&D activity	Inside, as a qualifying indirect activity	para 31(g)
Design, build and testing that resolve the uncertainty	Inside	para 27(c)
Testing after the final design is fixed and tested	Outside	example G
User trials testing commercial appeal	Outside	example H
Production, distribution and rollout	Outside	para 28(c)
Routine fault fixing after launch	Outside	para 35

Testing that directly contributes to resolving the uncertainty is R&D, but once the final design has been fixed and tested, "any further testing would not be R&D" (CIRD81910, example G). Rollout sits outside because production and distribution, administration and general support services do not directly contribute to resolving the uncertainty (para 28). An adviser who claims the whole rollout is stretching the boundary, one of the signs of an adviser overclaiming.

What the HMRC Manual Says: The Project Definition and Its Boundaries

HMRC reproduces the guidelines in full in its R&D tax relief manual at CIRD81910. Paragraph 19 states that "it is important to get the boundaries of the project correct", and that a project "could include a number of different sub-projects".

An activity is inside when it "must attempt to resolve an element of the scientific or technological uncertainty" (para 26). Qualifying indirect activities, such as R&D administration and training that directly supports the project, form part of a project without directly contributing (para 31) and still count as R&D (para 5). The R&D projects examples at the end of the guidelines are illustrative and "should be read in that context".

The Law and Guidelines Behind the Boundary

The R&D tax relief legislation defines research and development as "activities that fall to be treated as research and development in accordance with generally accepted accounting practice" (Income Tax Act 2007, s1006(2)). The Treasury may specify by regulations what is or is not R&D, "by reference to guidelines issued by the Secretary of State" (s1006(3) and (4)).

The current guidelines were introduced by Statutory Instrument 293 of 2023 for accounting periods beginning after 31 March 2023; earlier periods use the 2004 guidelines at CIRD81900. The guidelines are "given legal force by Parliamentary Regulations" (para 1), so the start and end rules in paras 33 and 34 have that force too.

Worked Example: Project Boundaries in Practice

The first of two R&D projects examples is the guidelines' spark plug illustration (CIRD81910, example J):

1. A company designs prototype plugs for an existing petrol engine and tests them in the engine. This is R&D (J1).
2. The uncertainty is resolved "once prototype plugs have been fully tested in the engine" (J1), and the plug work ends there.
3. The company then designs a new engine around the plugs, with a new combustion chamber design and lighter materials. This work has its own uncertainty and is R&D until "a functionally final prototype has been tested" (J2).
4. Producing plugs and engines for sale sits outside (para 28(c)).

A software project has the same shape. Suppose a company's engineers judge that no published method lets its app merge offline edits from many devices without losing data. Gathering user requirements is outside (para 33). The project starts when work on the merge method begins, includes the prototypes and tests that show whether merges hold (para 27(c)), and ends when a build with all the functional characteristics of the final engine passes its tests (para 34). Onboarding customers sits outside (para 28(c)).

How the Project Boundary Shapes Your R&D Tax Relief Claim

The additional information form asks for a global count of R&D projects in the accounting period and, for each project, its name and main field, the existing knowledge it planned to improve, the advance it aimed to achieve, the uncertainties it faced, how it sought to overcome them and its qualifying expenditure (HMRC manual CIRD182000, last updated 25 August 2026). HMRC states that "where businesses fail to answer the questions satisfactorily, this will result in their claim being rejected as invalid". How to write each answer is covered in what a weak R&D technical narrative looks like to HMRC.

Only costs of activity between the start and end points belong in the claim; the costs companies routinely leave out of an R&D claim are covered separately. Record these facts for each project as the work happens, using the guide to R&D record keeping:

• Uncertainty: the scientific or technological question the project set out to resolve.
• Start: the date work on that question began, with the record that shows it.
• End: the date it was resolved or the work stopped, and which.
• Testing and rollout: one line each, marked inside or outside.

Estimate Your R&D Tax Relief

GOV.UK's project checker gives HMRC's view on whether a project involves R&D, based on the information the company provides (GOV.UK, "Check if a project includes activities that qualify as Research and Development for tax purposes", 30 September 2025). The page adds that a qualifying project "does not always mean you can claim the tax relief on all R&D related costs", and the checker does not set start and end dates.

An R&D tax relief calculator works from costs, so its result depends on the boundary behind them. Costs from rollout or market testing inflate the estimate, and qualifying planning work left out shrinks it.

Conclusion

Write down each project's uncertainty, start date and end date, then run the R&D claim estimator with only the costs incurred between those dates to see what the claim is worth.