Key Takeaways
4 insights · 12 min readQualifying R&D activities must satisfy all five Frascati tests at once — novel, creative, uncertain, systematic, transferable or reproducible. Four out of five is a failed claim.
The uncertainty test is the one that decides most software claims: could a competent professional in the field have predicted the outcome using existing knowledge?
Only the portion of work carried out in the UAE qualifies. Offshore development is stripped out of the spend — and, it follows, out of the staff average that sets your rate band.
Social sciences, humanities and the arts are excluded absolutely, however novel or systematic the work.
A qualifying R&D activity under UAE corporate tax is work conducted in the UAE as part of an R&D project that is novel, creative, uncertain in outcome, systematic in conduct, and produces transferable or reproducible results. Eligibility is assessed against the OECD Frascati Manual, and all five conditions must be met together.
In this guide
What a Qualifying R&D Activity is The five tests The Frascati Manual Does software qualify? R&D partly outside the UAE Excluded fields Quick reference table Evidence for each test Why it matters for pre-approval Eligibility and the creditThe UAE R&D tax credit is generous enough that the interesting question is not how much you can claim but whether your work counts at all. Qualifying R&D activities are defined by reference to five conditions in Article 3 of Ministerial Decision No. 24 of 2026, and those conditions are lifted from the OECD Frascati Manual — the same framework the UK, Ireland, Singapore and most of the OECD use. They are conjunctive: fail one and the activity is out, however comfortably it clears the other four. This guide takes each test in turn with concrete pass and fail examples, then covers the three questions that decide most real cases: software, offshore work, and excluded fields. Once you know what qualifies, our guide to calculating the 15%, 35% and 50% credit shows what it is worth.
What is a Qualifying R&D Activity under UAE corporate tax law?
A Qualifying R&D Activity is an activity carried out within the UAE, as part of an R&D project, that satisfies all five conditions in Article 3 of Ministerial Decision No. 24 of 2026. Assessment is made by reference to the OECD Frascati Manual.
Two features of that definition do most of the work in practice. First, the unit of assessment is the activity within a project, not the company and not the department. A business can have an R&D team where only part of what the team does qualifies, and a business with no R&D department at all can have qualifying activity buried inside an engineering or production function. Mapping activities to projects, and projects to the tests, is the first task — not the last.
Second, the tests are conjunctive. This is not a weighted assessment where strength on novelty compensates for weakness on uncertainty. Each of the five is a gate. In practice claims fail on one of two: uncertainty, because the work was skilled but predictable; or systematic conduct, because nobody wrote anything down at the time.
What are the five tests for qualifying R&D activities?
The five criteria below are the Frascati Manual’s definition of research and experimental development, adopted into the UAE regime. Each has a distinct question behind it.
Test 1 — Novel: does it aim to produce new findings?
The activity must aim to generate knowledge that does not already exist. Novelty is measured against the existing stock of knowledge in the field, not against what your business happened to know before it started. Applying a well-understood technique in a new commercial setting is not novel simply because it is new to you.
One nuance worth knowing: where relevant knowledge exists but is not publicly available or readily deducible — held as a trade secret, or locked inside a proprietary system — independently working it out can still involve genuine advance. That argument has to be evidenced rather than asserted, and it is fact-specific.
✅ Likely qualifies
- Developing a polymer compound with properties not documented in the literature
- Establishing whether a known catalyst works at a temperature range nobody has characterised
❌ Likely does not
- Implementing an existing ERP system configured to your workflows
- Adopting a published manufacturing method already in use elsewhere in your sector
Test 2 — Creative: is it based on original concepts or hypotheses?
The work must rest on original ideas, concepts or hypotheses rather than the mechanical application of what is already known. This is the criterion that separates research from skilled technical execution — and skilled technical execution can be very sophisticated and still fail it. Formulating a new model, theory or approach passes; running an established procedure competently does not.
Test 3 — Uncertain: is the outcome unknown in advance?
There must be genuine scientific or technological uncertainty: the outcome, or the route to it, cannot be determined in advance by a competent professional in the relevant field working from existing knowledge. That framing matters. The test is not whether you were unsure; it is whether the uncertainty is objective.
This is the most frequently misapplied criterion in every R&D regime, and the UAE will be no different. Commercial uncertainty — will customers buy it, will it come in on budget — is not technological uncertainty. Difficulty is not uncertainty either: a large, complex build using documented tools and established methods is hard work, not research.
✅ Likely qualifies
- Testing whether a novel protein structure can be synthesised under conditions existing chemistry cannot predict
- Determining whether a process can hold tolerance at a scale never attempted
❌ Likely does not
- A complex website or app built entirely with documented frameworks
- Scaling a process that a competent engineer would expect to work
Test 4 — Systematic: is it planned, budgeted and recorded?
The activity must be conducted in a structured way, with a defined plan, stated objectives and an allocated budget. Ad hoc experimentation fails this test even when it produces a genuine discovery — the useful result of unplanned tinkering is still unplanned tinkering.
This is the test that ties directly to the documentation obligation in Article 12, which requires written, visual and electronic records of objectives, processes, methodologies, experiments and findings, retained for seven years. It is also the only one of the five that cannot be repaired after the fact. Novelty and uncertainty can be argued from technical evidence assembled later. Systematic conduct either was documented as it happened or it was not.
Test 5 — Transferable or reproducible: can the results be used elsewhere?
The results must be capable of being transferred or reproduced — the knowledge generated has to be recordable in a form others could use, and the work has to be repeatable rather than a one-off accident.
Here it is worth being precise, because this test is often over-read. The requirement attaches to the knowledge produced, not to the commercial deliverable. Bespoke client work is not automatically disqualified: if solving a client’s problem required a technological advance that your team could now apply to other clients or other problems, the underlying knowledge is transferable even though the deliverable is not. What fails is work whose only output is a customer-specific artefact with no generalisable learning behind it — which describes most contract engineering, but not all of it.
Not sure whether your project clears the uncertainty test?
Send us the technical brief. We will map the work against all five criteria and tell you where the gaps are before you spend anything on a pre-approval application.
What is the Frascati Manual and why does it govern eligibility?
The Frascati Manual is the OECD’s standard for defining and measuring research and experimental development. Article 3 of Ministerial Decision No. 24 of 2026 directs that eligibility be assessed by reference to it, which means UAE claims are being judged against an international framework with decades of interpretive material behind it — a considerable advantage when building a case.
Frascati divides R&D into three categories, and all three can produce qualifying activity:
| Category | What it is | Typical UAE example |
|---|---|---|
| Basic research | Experimental or theoretical work to acquire new knowledge, without a specific application in view | University-linked materials science with no immediate product target |
| Applied research | Original investigation directed at a specific practical aim | Investigating whether a desalination membrane chemistry can tolerate Gulf salinity |
| Experimental development | Systematic work drawing on existing knowledge to produce new or substantially improved products, processes or services | Developing a battery thermal management system for 50°C ambient operation |
Most commercial claims sit in the third category, and that is where the boundary with ordinary product development runs. Frascati’s own guidance is that development ceases to be R&D once the technological uncertainty has been resolved — so pilot production, market launch, tooling and routine debugging fall outside, even when they belong to the same project.
Does software development qualify as R&D in the UAE?
Sometimes — and the answer turns almost entirely on Test 3. Software is the single largest source of borderline claims in every R&D regime because the work is technical, valuable and difficult, and none of those three things is the test. The question is whether a competent software professional could have predicted the outcome from existing knowledge.
| Software work | Likely treatment | Deciding factor |
|---|---|---|
| New algorithm where no known approach solves the problem | Likely qualifies | Genuine technological uncertainty |
| Novel model architecture tested against a hypothesis | Likely qualifies | Original concept plus uncertain outcome |
| Achieving performance nobody has demonstrated at that scale | Likely qualifies | Outcome not deducible in advance |
| Building a complex platform with documented frameworks | Likely does not | Difficulty is not uncertainty |
| Integrating third-party APIs and services | Likely does not | Established methods, predictable outcome |
| Fine-tuning an existing model on your own dataset | Usually does not | Documented technique applied to new data |
| UI, UX and design work | Excluded | Falls within the arts exclusion |
The practical discipline for software teams is to isolate the uncertain component. A twelve-month platform build is not a qualifying activity, but the eight weeks the team spent resolving a genuinely unsolved concurrency or accuracy problem inside it may be. That requires project records granular enough to separate the two — which is a decision about how you run engineering, taken at the start of the year, not a decision about tax taken at the end of it.
What happens to R&D conducted partly outside the UAE?
Only the portion carried out within the UAE can be a Qualifying R&D Activity. Where a project runs across UAE and offshore teams, the offshore work is stripped out — same project, same team, same objective, but outside the credit.
⚠️ Offshore staff also cost you rate band, not just spend
The staff test that sets your credit rate counts R&D staff engaged in qualifying activity — and qualifying activity is UAE activity. It follows that offshore developers do not count toward the average, so an apparently large R&D team can fall short of the 6 or 14-staff thresholds once it is filtered to UAE personnel. Confirm the treatment for your facts before relying on it. See how the staff thresholds set the rate →
Worked example: the same project, apportioned
A Dubai technology company runs a development project with a team of 10 — 6 in the UAE and 4 offshore. Total project cost is AED 3,000,000, of which AED 1,850,000 relates to UAE activity.
| Position | As structured | If the whole team were UAE-based |
|---|---|---|
| Qualifying expenditure | AED 1,850,000 | AED 3,000,000 |
| Average R&D staff counted | 6 | 10 |
| Bands reached | 15% then 35% | 15%, 35%, then 35% on the balance |
| Credit | AED 447,500 | AED 850,000 |
The offshore arrangement costs AED 402,500 of credit in a single year — roughly a fifth of the project’s total cost. That does not mean offshore development is wrong; it means the credit belongs in the location decision rather than being discovered afterwards. Where a hybrid model stays, project-level cost and time allocation between UAE and non-UAE work has to be built into the accounting system from the start. Our accounting team sets up the cost centres and our payroll team maintains the headcount records that support the split.
Which fields are excluded from qualifying R&D activities?
Article 3 of Ministerial Decision No. 24 of 2026 excludes R&D conducted in three fields outright. The exclusion is absolute — work in these areas cannot qualify however novel, systematic or uncertain it is.
- Social sciences — economics research, psychology studies, market research, political science, sociology.
- Humanities — history, linguistics, philosophy, theology, cultural studies.
- The arts — design research, artistic experimentation, creative media.
The line to watch is a technical project with a social-science or design component. A behavioural study running alongside a medical device development is R&D in a social science field; the device engineering is not. Similarly, the visual design of a product interface sits in the arts while the engine behind it may not. The costs need separating at project level, and where a component is genuinely inseparable from an excluded field, expect it to follow the exclusion rather than the technical work.
Which activities qualify and which do not?
A quick reference for common UAE sectors. Treat it as a starting point rather than an answer — every line below is fact-specific, and the "depends" entries are where advice earns its fee.
| Activity | Likely outcome | Deciding factor |
|---|---|---|
| Pre-clinical pharmaceutical development | Qualifies | Novel, uncertain, systematic by design |
| New materials and chemical compound research | Qualifies | Strong fit across all five criteria |
| Experimental energy storage technology | Qualifies | Uncertain outcome, transferable findings |
| Agricultural biotechnology trials | Qualifies | Novel, creative, reproducible |
| Desalination or water treatment process development | Qualifies | Applied research with genuine technical unknowns |
| Novel algorithm development with unsolved problem | Depends | Must clear the uncertainty test on the facts |
| AI model training on a novel problem | Depends | Novel architecture qualifies; fine-tuning usually does not |
| Routine software development on known frameworks | Does not qualify | No technological uncertainty |
| Process optimisation using established methods | Does not qualify | Not novel, outcome predictable |
| Quality control and production testing | Does not qualify | Known outcome, not experimental |
| Market research and consumer surveys | Does not qualify | Social sciences exclusion |
| Product and interface design work | Does not qualify | Arts exclusion |
What evidence do you need to prove each test?
Article 12 requires technical documentation covering objectives, processes, methodologies, experiments and findings, retained for seven years from the end of the relevant tax period. That is the obligation. The more useful question is which records prove which test, because a file that satisfies Article 12 in volume can still leave a specific criterion unevidenced.
| Test | What proves it |
|---|---|
| Novel | Literature and patent searches, prior-art review, a statement of the knowledge gap the project targets |
| Creative | Hypothesis statements, design rationale, records of approaches considered and rejected |
| Uncertain | Technical narrative naming the specific unknown, failed experiments, iteration logs, competent-professional sign-off |
| Systematic | Dated project plan, defined objectives and milestones, approved budget, experiment records made at the time |
| Transferable or reproducible | Method documentation, results written up in a usable form, evidence of reuse or replication |
Note how much of this is generated during the work rather than around it. Failed experiments are among the strongest evidence you can hold for the uncertainty test — teams that only record what worked throw away their best material. A short technical narrative written at the start of each project, naming the unknown and the approach, does more for a claim than a thick file assembled twelve months later.
Why does eligibility matter for pre-approval?
Pre-approval is a mandatory gateway: the credit cannot be claimed on a corporate tax return without it. The reviewing body assesses your project documentation against the five criteria, so a project that fails one test — or that cannot evidence systematic conduct through contemporaneous records — does not get through.
The sequencing consequence is the whole point of this guide. Eligibility is not a filing-season question. By the time a tax period has closed, the systematic test is already decided, the failed experiments have or have not been recorded, and the UAE-versus-offshore time split either exists in the timesheets or does not. Businesses that treat R&D eligibility as an accounting exercise at year end tend to discover that the technical case they needed was available in March and gone by January.
The practical model is to run a short eligibility assessment at project initiation, document the technical narrative while the uncertainty is live, and keep the cost allocation running monthly. That is cheaper than a rejected application and considerably cheaper than an approved claim that does not survive later review.
How do eligibility and the credit calculation fit together?
Eligibility determines what goes into the claim; the calculation rules determine what it is worth. The two interact more than people expect, and the interaction runs through headcount.
Qualifying expenditure is credited in bands — 15% on the first AED 1,000,000, 35% on the next AED 1,000,000, and 50% from AED 2,000,001 up to the AED 5,000,000 cap — with each band also requiring a minimum average of 2, 6 and 14 R&D staff respectively. Because both the spend and the staff count are filtered by what qualifies, a narrow eligibility position compresses the claim twice: less expenditure, and fewer counted staff, which can drop you a rate band on top. The reverse also holds — identifying qualifying activity inside an engineering function you never thought of as R&D can lift both figures at once. Full mechanics, with worked examples and the real AED 2,000,000 annual ceiling, are in our R&D tax credit calculation guide.
One structural point to settle early if you are a free zone company: claiming the credit and then electing Qualifying Free Zone Person status within five years triggers claw-back, as does applying Small Business Relief. The R&D credit and the 0% free zone route are alternatives, not a stack — our guides to processing and headquarter services set out what the QFZP route demands in return.
📅 Small Business Relief closes on 31 December 2029 — and the election is annual
Small Business Relief is available only for tax periods ending on or before 31 December 2029. It must be elected in the corporate tax return for each eligible tax period separately, it cannot be backdated once the return is filed, and a missed election forfeits that year’s relief. Separately, once revenue exceeds AED 3,000,000 in any tax period, relief ends for that period and for all subsequent tax periods, even if revenue later falls back below the threshold. Because electing SBR within five years also claws back an R&D credit, an early-stage research company sitting under the threshold should model both routes before electing either. Governed by Ministerial Decision No. 73 of 2023. Check your SBR eligibility →
Key terms used in this guide
| Term | What it means |
|---|---|
| Qualifying R&D Activity | Activity conducted in the UAE as part of an R&D project that meets all five Article 3 conditions |
| Frascati Manual | The OECD standard for defining and measuring research and experimental development, used to assess eligibility |
| Technological uncertainty | Uncertainty a competent professional in the field could not resolve from existing knowledge — not commercial or budgetary risk |
| Experimental development | Systematic work using existing knowledge to create new or substantially improved products, processes or services |
| Conjunctive test | All conditions must be satisfied together; strength on one does not offset failure on another |
| Pre-approval | Mandatory authorisation obtained before the credit is claimed in the corporate tax return |
| Small Business Relief | Elective relief under MD 73/2023 for revenue up to AED 3,000,000, for tax periods ending on or before 31 December 2029; electing it within five years claws back the R&D credit |
| Contemporaneous records | Documentation created while the work happens — the only evidence that satisfies the systematic test |
Most businesses that ask whether they do R&D are asking the wrong question. The right one is which specific activities, inside which projects, resolved an uncertainty a competent professional could not have resolved from what was already known — and whether anyone wrote it down at the time. Answer that and the rest is arithmetic. Our corporate tax team runs the eligibility assessment, builds the technical narrative and manages pre-approval end to end. Wider context: the UAE corporate tax guide for businesses.
Fastlane Tax Team
FTA-registered tax agents and Ministry of Economy registered auditors with 4,000+ corporate tax and VAT filings across the UAE mainland and 40+ free zones. Every guide is reviewed against current FTA regulations before publishing.
Ask the team a question