TLDR: R&D advisory services cover eligibility assessment, core and supporting activity drafting, evidence strategy, AusIndustry registration and expenditure calculation for the R&D Tax Incentive. They are optional: eligibility is self-assessed, and companies can prepare their own claim. What you are really buying is judgement on activity boundaries, exclusions and evidence, so price the engagement against that, not against the size of the refund.
What R&D advisory services cover
R&D advisory is the professional support layer around a self-assessed tax program. Nothing in Division 355 ITAA 1997 requires an advisor. The company remains responsible for the registration it signs and for the R&D tax schedule its tax agent lodges.
Most engagements cover four areas: deciding which activities meet the core R&D test, writing them up for AusIndustry, calculating notional R&D expenditure, and preparing the evidence file that supports both.
Note the split in administration. Activity eligibility is decided by Industry Innovation and Science Australia and its delegates at AusIndustry. Expenditure and the offset are administered by the ATO. An advisor who only understands one side leaves a gap.
The advisory workflow, step by step
Most engagements follow the same sequence, whether the provider is a Big Four team or a platform.
- Scoping. Identify candidate projects and, within them, discrete experimental activities. Eligibility is assessed per activity, never per project or per product.
- Eligibility assessment. Test each candidate against s 355-25(1): outcome unknown in advance to a competent professional anywhere in the world, systematic progression from hypothesis to experiment, observation and evaluation, and a purpose of generating new knowledge.
- Exclusion screening. Check the s 355-25(2) list, including internal administration software at paragraph (h) and market research at paragraph (a).
- Evidence review. Collect contemporaneous records: repositories, issue trackers, test results, dated hypothesis notes.
- Drafting. Write each registration field to its own requirement.
- Expenditure calculation. Apportion salaries, contractor payments, overheads and decline in value by R&D time.
- Registration and handover. Submit the registration to AusIndustry on the company's behalf (the company remains responsible for what is registered), then hand the expenditure figures to the tax agent for the ATO schedule.
Registration fields and character minimums
Step 5 is where competence shows. The AusIndustry registration, accessed via the customer portal linked from business.gov.au, is not one narrative: each core activity is split across separate fields with their own character minimums.
| Field | Question | Minimum characters |
|---|---|---|
| Sources investigated | Q79 | No stated minimum |
| Why a competent professional could not know the outcome | Q80 | 600 |
| Hypothesis | Q82 | 650 |
| Experiment (the method) | Q83 | 650 |
| Evaluation of results | Q84 | 530 |
| Conclusions reached in the income period | Q85 | 340 |
| New knowledge produced | Q89 | 600 |
Types of provider in Australia
The Australian market splits into four broad provider types: Big Four and mid-tier accounting firms, specialist R&D boutiques, generalist accountants, and software-led platforms. Each trades depth in one area for gaps in another.
| Provider type | Typical strength | Typical weakness |
|---|---|---|
| Big Four and mid-tier accounting | Complex group structures, large claims, integration with the tax return | Junior-staffed drafting, generic technical narratives |
| Specialist R&D boutiques | Deep program knowledge, sector focus, review experience | Capacity constraints, variable technical depth outside their niche |
| Generalist accountants | Already hold your financials | Often treat the registration as a form-filling exercise |
| Software-led platforms | Evidence pulled from engineering systems, per-field drafting at scale | Less suited to unusual structures or heavy grant interactions |
There is no regulated category called "R&D advisor", so anyone can use the title, but providing tax agent services for a fee, including advice on your R&D tax position, requires registration as a tax agent.
Ask about registration status directly, because it also determines who can lodge the ATO schedule.
Pricing models
Pricing usually follows one of three models: a success fee as a percentage of the offset, a fixed fee, or hourly rates. Success fees of 10% to 20% remain common in traditional consulting. Fixed fees start at a few thousand dollars for a small single-activity claim.
The problem with percentage pricing is that the fee scales with your payroll, not with the work.
A worked example
A software company with aggregated turnover under $20M and $500,000 of notional R&D expenditure claims the refundable offset at the company tax rate plus 18.5 percentage points, so 43.5% for a base-rate entity at 25%. That is a $217,500 offset.
| Fee model | Cost | Net benefit |
|---|---|---|
| Success fee at 20% | $43,500 | $174,000 |
| Success fee at 15% | $32,625 | $184,875 |
| Success fee at 10% | $21,750 | $195,750 |
| Fixed fee of $8,000 | $8,000 | $209,500 |
The registration work is materially identical in all four cases. A fuller breakdown sits in our consultant versus software cost comparison.
Also confirm the minimum spend rule applies to you: at least $20,000 of notional R&D deductions is required unless the expenditure goes to a Registered Research Service Provider or is a Cooperative Research Centre monetary contribution.
What to look for in an advisor
Judge on method, not on claimed success rates. The signals below separate a defensible engagement from an expensive one.
They ask for technical evidence first
If the intake list is a payroll export and a product roadmap, the registration narrative will be reconstructed after the fact. In the Absolute Vision tribunal decision, the gap between the registration narrative and the records on file was fatal. Evidence habits are covered further in our contemporaneous evidence guide.
They write real hypotheses
A hypothesis must be falsifiable and carry a measurable target under defined conditions. "We wanted to see if our approach was more scalable" is a business objective.
Two other tribunal decisions, GQHC and Active Sports, both turned on the quality of the hypothesis, and in GQHC on whether a genuine evaluation step existed. Evaluation is the element tribunals most often find missing.
Clinical uncertainty, not just technical
For a health-tech company, the eligible unknown is frequently whether the intervention shifts a clinical endpoint, not whether the app can be built. A structured study measuring a pre-specified clinical or behavioural endpoint is not market research, even when it runs on A/B infrastructure. The discriminator is the pre-stated hypothesis and endpoint.
Findings come up early in the conversation
Advance findings must be applied for before the end of the income year the activities are conducted in. Overseas findings must be applied for before the end of the first income year the overseas activities are conducted, and there are no extensions under any circumstances.
An advisor engaged in March cannot fix an overseas finding that should have been applied for by the end of a previous income year. A finding for overseas work first conducted this year can still be lodged before the end of the current income year (30 June for a standard year end).
Questions to ask before engaging
Ask these before signing, and get the answers in writing.
- Who drafts the registration, and what is their technical background?
- Will you interview our engineers or clinicians directly, or work from a template?
- What is the total fee, including any success component, and what triggers it?
- Is support during an AusIndustry or ATO review included, or billed separately?
- Are you a registered tax agent, and if not, who lodges the R&D schedule?
- What contemporaneous evidence do you require from us, and by when?
- How do you handle activities that may fall inside the s 355-25(2) exclusions?
Any answer that promises approval is a reason to walk. Nobody can guarantee an outcome that IISA and the ATO decide after the fact, with review rights running to the Administrative Review Tribunal.
Where poor engagements go wrong
The same failures come up again and again in reviews.
- Over-claiming: whole projects registered as core activities, with routine integration and configuration work swept in. The core versus supporting distinction exists for a reason.
- Under-claiming: genuine experimental work left out because the advisor did not understand the domain, or because a clinical endpoint study was mistaken for market testing.
- Narrative-first drafting: eloquent registrations with no records behind them. This is the profile that fails an ATO review.
- Deadline drift: registration is due 10 months after year end, so 30 April for a 30 June year end. It is a hard statutory deadline and it must precede claiming the offset. See our deadlines guide and what triggers ATO reviews.
Software-led advisory compared
The structural difference is where the source material comes from. Traditional consulting reconstructs the year from interviews near the deadline. Software-led advisory reads the engineering record as the work happens: commits, pull requests, issue histories and test runs, mapped to activities.
Rand works this way. It builds the claim from GitHub and Jira evidence, the company's team reviews it, and Expert Reviewers approve the AusIndustry registration. Pricing is a service fee, not a percentage.
The trade-off is judgement-heavy edge cases. Complex group structures, grant recoupment interactions, feedstock adjustments and overseas findings still benefit from a specialist conversation.
Who actually needs advisory support
Not everyone. A single-activity claim, with clear technical uncertainty and a team that already writes design docs and test results, is genuinely self-preparable using the end-to-end guide and the AusIndustry field guidance.
Support earns its fee where the facts are contested: activities near an exclusion boundary, overseas work, associate payments that must be paid in cash by year end, government grants that trigger clawback, or a first claim where the evidence habits do not yet exist.
The right question is not "should I get help" but "which specific judgement calls in my claim need it". Answer that first, then price the engagement against it. Rand exists for companies whose R&D is already recorded in their engineering systems and who would rather not pay a percentage of the refund to have it retyped.