TLDR: A digital health company's study measuring a clinical or behavioural outcome can be eligible R&D under the R&D Tax Incentive, even where the app itself is ordinary to build, provided the study meets the four-part test in s 355-25(1) ITAA 1997. The core activity is the outcome experiment, with the build and data pipeline registered as supporting activities under s 355-30.
The line that decides most of these claims is the one between a pre-specified clinical or behavioural endpoint and excluded market research.
Where the eligible uncertainty sits in a digital health app
In most health-tech companies the uncertainty is not in the code. Building a CBT module, a symptom diary or a coaching sequence uses established patterns, and a competent software engineer can predict the outcome. What nobody can predict is whether the intervention shifts a measured health outcome.
That second question can be a genuine knowledge gap. Whether a six-week app-delivered program reduces IBS symptom severity scores, or whether self-guided delivery preserves the effect sizes reported for clinician-delivered protocols, may not be answerable from the published literature. Where your documented worldwide search shows it is not, the question can only be resolved by running the study.
So the registration should follow the uncertainty. If the clinical endpoint is the unknown, the core activity is the study, and the build is supporting work. Claiming the reverse (dressing up routine product engineering as the experiment) is the fastest way to lose the claim.
Some health-tech companies have both streams: a genuine technical hurdle (an on-device inference model holding accuracy at a stated latency, for example, see software R&D Tax Incentive eligibility) plus a clinical or behavioural unknown. Register them as separate core activities. See core vs supporting R&D activities for how the split works in practice.
The statutory test applied to a clinical outcome study
The test does not change by sector. A clinical outcome study is assessed against the same provision a software company applies to an algorithm, and AusIndustry's guidance on the R&D Tax Incentive applies the same criteria whatever the field.
Section 355-25(1) requires:
- An experimental activity whose outcome cannot be known or determined in advance on the basis of current knowledge, information or experience.
- An outcome that can only be determined by applying a systematic progression of work based on principles of established science.
- Work that proceeds from hypothesis to experiment, observation and evaluation, and leads to logical conclusions.
- Activity conducted for the purpose of generating new knowledge, including new knowledge in the form of new or improved materials, products, devices, processes or services.
All four elements must be satisfied. The legislation groups them into two limbs: items 1 to 3 unpack limb (a), the outcome that cannot be known in advance, and item 4 is limb (b), the new-knowledge purpose.
For a clinical stream, the "established science" is clinical and behavioural science, and the systematic progression is your protocol.
Limb (a): outcome not knowable in advance
The benchmark is a competent professional with access to knowledge that is publicly available or reasonably accessible anywhere in the world. For a clinical stream, that professional is a clinician or health researcher in the relevant condition area, not your lead engineer.
The question to answer is precise: what could that clinician not predict, and why. A strong answer looks like this. Published efficacy evidence exists only for clinician-delivered protocols, and a competent clinician could not predict whether the effect size survives self-guided app delivery, because delivery format changes dose, adherence and the therapeutic relationship.
"New to our company" never satisfies this limb. In Absolute Vision Technologies [2022] AATA 2319 the Tribunal noted the absence of any literature review to confirm that a knowledge gap existed at all.
Your worldwide search of the clinical literature, trial registries such as ANZCTR and ClinicalTrials.gov, and published adherence data must predate the study and be documented.
Limb (b): purpose of generating new knowledge
The purpose must be to generate new knowledge, including new knowledge in the form of new or improved materials, products, devices, processes or services. A study whose purpose is to produce a marketing claim, or to decide pricing, does not satisfy this limb.
State the knowledge you set out to create, not the commercial benefit. For example: establishing the first effect-size and adherence profile for self-guided delivery of a protocol in a defined population, and characterising which session-completion patterns predict response.
Is an in-app outcome study market research?
Usually not, if it genuinely tests a clinical or behavioural endpoint against a pre-stated hypothesis. Section 355-25(2)(a) excludes market research, market testing, market development and sales promotion, including consumer surveys, from ever being a core activity, but no exclusion attaches to the tooling.
In practice that means one A/B platform can host an eligible adherence experiment and an excluded conversion test in the same quarter. Keep their protocols, endpoint definitions and results in separate records, because a reviewer reads the endpoint before anything else.
The discriminator: pre-specified endpoint and hypothesis
What decides the question is the endpoint you pre-specified and the hypothesis you documented before running the experiment.
Measure symptom change, program completion, dose delivered or another behavioural endpoint tied to therapeutic effect, and the study is usually outside the exclusion. Measure trial-to-paid conversion, churn or revenue, and you are squarely inside it. Where an experiment collects both, what it was designed to answer, shown by the pre-registered primary endpoint, decides which side it falls on.
A second exclusion also needs checking. Research in the social sciences cannot be a core activity (s 355-25(2)(d)), and AusIndustry lists psychology among the subjects it is likely to cover. What decides it is the substance of the experiment: a reminder test built to deliver a protocol's dose is a question about the treatment, while the same test run to learn why people open an app is a question about behaviour, and adding a symptom score afterwards does not change that. In Body by Michael [2025] ARTA 44 the Tribunal said, in obiter, that health-focused work was not social science, but the claim still failed because the outcome could be known.
This also means retrospective relabelling fails. Deciding in April that last September's onboarding test was really an adherence study will not survive the contemporaneous evidence requirement.
Examples that fall on each side of the line
The exclusions in s 355-25(2) bite on the endpoint and the purpose, not the format. These examples show how the categories apply to typical digital health work.
| Activity | Treatment under s 355-25(2) |
|---|---|
| Randomised test of reminder cadence with a pre-stated 80% session-completion endpoint | Can be a core activity (adherence is the dose) |
| Same test measured on subscription conversion | Excluded market research, s 355-25(2)(a) |
| Single-arm pilot with a protocol-defined symptom-score primary endpoint | Can be a core activity |
| Willingness-to-pay survey or product-market-fit interviews | Excluded market research |
| Usability testing with no behavioural or clinical endpoint | Not core; may be supporting if it enables the experiment |
| Preparing a TGA submission or ISO 13485 certification pack | Excluded as core under s 355-25(2)(f); can only ever be supporting, and must pass the dominant purpose test |
The last row is a compliance exclusion rather than market research. Being regulated does not itself trigger it: work undertaken to show compliance with a standard is excluded, while genuine experimentation that happens to occur in a regulated environment is not.
The clinical trials determination, and when it does not apply
The shortcut only reaches trials of a therapeutic good that is not yet on the ARTG.
Under the Industry Research and Development (Clinical Trials) Determination 2022, phase 0, I, II and III trials and pre-market pilot and pivotal stage trials of a therapeutic good not yet on the ARTG are treated as core R&D activities, provided the trial runs under the TGA's clinical trial notification (CTN) or approval (CTA) arrangements.
It does not extend to post-market phase IV trials, trials of generic products, or any market research, market testing or sales promotion component of a trial.
Where the determination applies, typically regulated software as a medical device being trialled pre-ARTG, it removes the need to argue the four-part test, but the registration form still asks for the trial phase, protocol details and the good's ARTG status.
Many wellness and lifestyle apps are not regulated as medical devices, and some digital mental health tools are excluded from TGA regulation under item 14E of the Excluded Goods Determination when they meet its conditions. Where the app being studied is not a therapeutic good, the determination is unavailable. The ordinary test in s 355-25(1) applies instead, and a genuine efficacy study can qualify under it. More detail in our guide to clinical trial R&D Tax Incentive eligibility.
Typical core activities for a digital health app
Divide the registration by hypothesis, never by product, sprint or funding round. Most registrations carry three to seven core activities.
- Efficacy study. Does the intervention move a defined clinical score by a pre-stated margin in a defined population over a defined period?
- Adherence or dose-delivery experiment. What sustains completion at the level required for therapeutic effect? Adherence to a specific digital intervention in a specific population is often unresolved; show it with your own documented search.
- Personalisation or triage logic. Does an algorithmic assignment rule produce better outcomes than a fixed protocol, where the response profile is not predictable from published data?
- Technical hurdle, where one exists. A measurable target (on-device inference latency, signal accuracy from consumer sensor data) at a scale or condition no published result demonstrates.
Write each one to its own hypothesis. AusIndustry's sector guidance warns against over-grouping: combining several unknowns into one core activity makes the description unclear and can affect eligibility. It also collides with GQHC [2024] AATA 409, which requires each hypothesis to be specific and capable of being validated or invalidated.
Supporting activities: the build, the data pipeline, the recruitment
Section 355-30 ITAA 1997 covers activities directly related to core R&D activities. For a clinical stream, that typically includes building the intervention module used in the study, instrumenting outcome capture, constructing the analysis dataset, participant recruitment and screening, and site or CRO coordination work.
CRO and site arrangements also raise the conducted-for and at-risk questions covered under common failure reasons below, so confirm who bears the financial risk before registering them.
Two constraints matter. First, "directly related" is a real test: general product engineering unconnected to the registered experiment does not qualify. Second, activities that fall into an excluded category, or that produce goods or services (or are directly related to producing them), must also satisfy a dominant purpose test.
Usability testing sits here rather than in core, and only where it directly enables an adherence or efficacy experiment.
Evidence AusIndustry and the ATO expect for a clinical endpoint claim
Contemporaneous records decide these claims. GQHC held that hypotheses must be scientific, specific and capable of being validated or invalidated, that observation and evaluation must be evidenced rather than asserted, and that maintaining adequate documentation is itself part of the systematic progression.
For a clinical endpoint claim, the expected evidence hierarchy is:
- Study or trial protocol, with the primary endpoint and dosing schedule stated
- HREC approval and correspondence, and CTN notifications or CTA approvals where applicable
- Statistical analysis plan with pre-registered endpoints and analysis rules
- Literature review notes and registry searches, dated before the study started
- De-identified participant outcome and adherence data
- Investigator or clinical advisory meeting minutes recording interim decisions
- CRO and site contracts, plus the usual engineering artefacts for the software components
Royal Wins Pty Ltd [2020] AATA 4320 makes the consequence plain: without contemporaneous documentation that a hypothesis was developed and tested, the activities are not eligible however much work was done.
The Tribunal in Active Sports Management treated the contemporaneous formation of the hypothesis as the essential starting point of the activities, not a write-up produced later. See contemporaneous evidence for a defensible R&D claim.
How this maps to the AusIndustry registration fields
The registration is structured as projects, then core activities, then supporting activities, lodged through the customer portal reachable from AusIndustry's R&D Tax Incentive page on business.gov.au. The current form version was released 15 August 2025.
A core activity is not one narrative: it is split across separate fields, each with its own character minimum.
| Registration field | What a clinical stream puts here |
|---|---|
| Sources investigated | Systematic reviews and meta-analyses consulted, trial registry searches, published adherence data, and the specific gap left |
| Why a competent professional could not know the outcome | Why a clinician could not predict the result, tied to what published evidence does and does not cover |
| Hypothesis | The pre-stated, testable proposition with thresholds (for example a defined score change and a completion rate) |
| Experiment | The method only: design, arms, cohort, what was varied and held constant, interim checkpoints |
| Evaluation | How measured results were analysed against the pre-stated thresholds, and how each evaluation drove the next iteration |
| New knowledge and conclusions | What the study established that was not previously demonstrated, including where the hypothesis was disproven |
Treat each field as a separate question with its own standard of proof. Our guide on documenting R&D activities for AusIndustry works through the field-level detail.
What the offset is worth to a pre-revenue health-tech company
If aggregated turnover is under $20M and the company is not controlled by income-tax-exempt entities, the offset is refundable at the company tax rate plus 18.5 percentage points, commonly 43.5% for a base-rate entity taxed at 25%. Refundable means it is paid as cash to the extent it exceeds tax payable.
Worked example
Suppose a pre-revenue digital therapeutics company records $850,000 of notional R&D deductions for FY2026: apportioned salaries for engineering and clinical staff on the registered activities, CRO and site costs, and directly related overheads.
| Item | Amount |
|---|---|
| Notional R&D deductions | $850,000 |
| Offset rate (25% + 18.5 points) | 43.5% |
| Refundable offset | $369,750 |
This assumes Australian sites, that the company bears the financial risk and holds rights to results, and that the integrity rules (at-risk, clawback, feedstock) do not reduce the claim. Overseas site costs need an Overseas Finding before they can be included.
Thresholds to keep in view
You need at least $20,000 of notional R&D deductions to access the offset. The floor does not apply where the spend is with a Registered Research Service Provider that is not an associate, or is a monetary contribution under the Cooperative Research Centre program.
At $20M or more of aggregated turnover the offset becomes non-refundable, at the company tax rate plus a two-tier intensity premium applied marginally.
| Aggregated turnover | Offset |
|---|---|
| Under $20M | Refundable: company tax rate + 18.5 points (commonly 43.5%) |
| $20M or more | Non-refundable: company tax rate + 8.5 points on notional R&D up to 2% of total expenditure, + 16.5 points on the portion above 2% |
The 2026-27 Budget proposes raising the $20M split to $50M from 1 July 2028. It is exposure draft legislation (Treasury consulted from 10 to 28 September 2026), not yet law, and it does not affect current income years. Current rate and expenditure rules are set out in the ATO's R&D Tax Incentive guidance.
Common reasons digital health claims fail
- The hypothesis is a commercial goal. "We wanted to see if users liked the new program" is not a hypothesis capable of being invalidated.
- No documented worldwide knowledge gap. No literature review, no registry search, no dated evidence of the enquiry.
- The endpoint is commercial. An experiment measured on conversion or retention is excluded market research, whatever it is called in the registration.
- Records are retrospective. Documents prepared after year end for an internal review are not contemporaneous, and inconsistency between records and the registered activities is fatal.
- Compliance work claimed as core. TGA dossier preparation, quality-management certification and post-market surveillance are excluded core activities under s 355-25(2)(f), and can only ever be supporting activities.
- Conducted-for and at-risk problems. Trials run through CROs, associated entities or for a foreign parent raise the question of who bears the financial risk and holds the rights to results. These are active ATO risk areas (Taxpayer Alerts TA 2023/4 and TA 2023/5) and should be surfaced, not drafted around.
Deadlines and the order of operations
Registration with AusIndustry is due 10 months after the end of the income year: 30 April for a 30 June year end. It is a statutory deadline: AusIndustry approves an extension of 14 days or less requested before it, and considers longer extensions (capped at 92 days) case by case. Registration must precede claiming the offset in the company's tax return.
The sequence that keeps a clinical claim defensible:
- Run the worldwide literature and registry search, and record it.
- Document the hypothesis, primary endpoint and analysis plan before the study starts.
- Run the study, capturing observations and interim evaluations as they happen.
- Register the core and supporting activities in the portal before the 10-month deadline.
- Have the tax agent lodge the R&D tax schedule with the company return.
Getting steps one and two in the right order is the whole game. See every R&D Tax Incentive deadline that matters for the full calendar.
Rand builds the registration from the evidence your team already produces, engineering artefacts alongside protocols and endpoint data, so the clinical and technical streams are drafted as separate core activities and finalised by an Expert Reviewer before your company declares and submits the registration. If your app's real unknown is whether it changes a health outcome, that is where the claim should be built.