R&D Tax Incentive for Software - Fixed Fee Advisory

R&D Tax Incentive for Software Companies – Prime Innovation

Does Your Software Development Qualify?

Australia’s R&D Tax Incentive is one of the most valuable government programs available to software companies – but the line between eligible R&D and routine development is where most claims succeed or fail. Not all coding is R&D. The program rewards companies that push beyond existing knowledge to solve genuine technical problems, not those simply building products using established tools and techniques.

For eligible companies with aggregated turnover under $20 million, the incentive provides a 43.5% refundable tax offset – meaning pre-revenue startups and loss-making companies receive a cash payment from the ATO. Larger companies (turnover $20-500 million) receive a non-refundable offset at their corporate tax rate plus an intensity premium.

Prime Innovation, a specialist division of Prime Partners, helps software companies identify eligible R&D activities, structure claims correctly and maintain the documentation that withstands AusIndustry scrutiny.

Discuss Your Software R&D Claim

What Software R&D Qualifies

The R&D Tax Incentive is governed by Division 355 of the Income Tax Assessment Act 1997. For software activities to qualify as core R&D activities, they must involve technical uncertainty, systematic investigation and the purpose of generating new knowledge.

Algorithm Development

Creating new algorithms to solve computational problems where no known solution exists or existing solutions are inadequate for the specific constraints.

AI & Machine Learning

Developing new model architectures, training methodologies or approaches where the feasibility of achieving target accuracy is uncertain.

Cybersecurity Research

Investigating new approaches to threat detection, encryption methods or vulnerability prevention where existing techniques are insufficient.

Distributed Systems

Designing systems to handle scale, consistency or fault tolerance challenges beyond what proven patterns can address.

NLP Research

Developing new approaches to language understanding, generation or analysis where accuracy or reliability of outcomes is uncertain.

Computer Vision

Creating new methods for image recognition, object detection or visual analysis involving unproven techniques.

Performance Engineering

Investigating techniques to achieve processing speeds or throughput that cannot be achieved through known optimisation methods.

Supporting Activities

Test environments, datasets, prototypes and documentation that directly support core R&D activities also qualify.

What Does Not Qualify

Understanding what falls outside the R&D Tax Incentive is equally important. The following software development activities are not eligible:

Routine coding – Building features using established frameworks and design patterns (React frontends, REST APIs, CRUD operations)
Bug fixes and maintenance – Fixing defects in existing software
UI/UX design – Visual design, interface changes, A/B testing of layouts
Deploying known frameworks – Installing and customising off-the-shelf platforms

The Grey Area

Many software activities sit in a grey area. Integrating a pre-trained ML model is generally not R&D. But fine-tuning that model for a novel domain where performance is uncertain – and systematically experimenting to achieve acceptable accuracy – may qualify.

Systems integration – Connecting systems using documented APIs
Standard DevOps – CI/CD pipelines, containerisation using established tools
Porting and migration – Moving applications between platforms using known methods
Commercial adaptation – Modifying open-source software without resolving technical uncertainty

The Technical Uncertainty Test for Software

Technical uncertainty is the cornerstone of every R&D claim. It exists when a competent professional in the relevant field cannot know or determine the outcome of an activity in advance. It is not sufficient that the outcome is unknown to your team – it must be unknown to any competent professional.

Scenario	Technical Uncertainty?	Why
Recommendation engine using collaborative filtering	No	Well-established technique
Recommendation engine on sparse, cold-start data at sub-50ms latency	Potentially yes	Combination of constraints may exceed known solutions
Implementing OAuth2 authentication	No	Standard protocol with documented implementations
Zero-knowledge proof authentication for a novel use case	Yes	Novel application requiring experimental validation
Training a GPT model on industry-specific data	Generally no	Fine-tuning pre-trained models is established practice
New training methodology for domain-specific LLMs with unacceptable hallucination rates	Yes	Requires investigation beyond current knowledge

Core vs Supporting Activities

Core Activities (Section 355-25)

Core activities are the experimental investigations themselves – the work that directly addresses the technical uncertainty:

Writing experimental code to test hypotheses
Designing and running performance benchmarks
Training and evaluating model architectures
Conducting systematic experiments on algorithm variants

Supporting Activities (Section 355-30)

Supporting activities enable or facilitate core R&D but do not themselves involve experimentation:

Setting up cloud infrastructure for R&D workloads
Building data pipelines for training datasets
Creating testing frameworks for experimental code
Writing technical specifications that guide experiments

Important: Supporting activities must have a direct, close and relatively immediate connection to specific core R&D activities.

Worked Example – Software Company R&D Claim

Company profile: SaaS company developing an AI-powered compliance platform. 8 developers, 2 data scientists. Aggregated turnover $3.5 million. R&D activities: developing novel NLP models for regulatory document analysis.

Category	Total Cost	R&D Eligible	Notes
Developer salaries (R&D team)	$400,000	$250,000	62.5% based on timesheets
Data scientist salaries	$300,000	$270,000	90% on core R&D
Cloud computing (AWS)	$120,000	$48,000	40% of total cloud spend
Contractor (ML specialist)	$80,000	$80,000	Exclusively for R&D
Software tools and licences	$40,000	$12,000	R&D portion only
Total	$940,000	$660,000

R&D Tax Incentive offset: Eligible expenditure of $660,000 at 43.5% = $287,100 refundable tax offset. Net cost of R&D after offset: $372,900 – effectively a 44% discount on R&D costs. Because turnover is under $20 million, the offset is refundable – the ATO pays $287,100 in cash regardless of profitability.

Common Mistakes Software Companies Make

1. Claiming All Development as R&D

Software companies often assume that because they are building something new (to them), all development qualifies. AusIndustry distinguishes between commercial novelty and technical novelty.

2. Insufficient Documentation

Software development moves fast and R&D documentation often falls behind. Without contemporaneous records of technical uncertainty, hypotheses tested and results observed, claims are difficult to defend.

3. Failing to Separate R&D from Production

When the same developers work on both R&D and production code, time must be accurately apportioned. Claiming 100% of a developer’s time as R&D will trigger scrutiny.

4. Overclaiming Cloud Costs

Cloud costs are only eligible to the extent they relate to R&D. Companies without proper tagging or separation between R&D and production workloads cannot substantiate their claims.

5. Missing the Registration Deadline

R&D activities must be registered with AusIndustry within 10 months of the end of the financial year. Missing this deadline means the claim is lost for that year.

6. Confusing Commercial Risk with Technical Uncertainty

Market risk (“will customers buy this?”) and technical uncertainty (“can we build this?”) are different things. Only technical uncertainty qualifies.

How to Document Software R&D

Git Commits & Version Control

Your git history is a contemporaneous record. Use descriptive commit messages referencing R&D hypotheses, tag experimental branches (e.g. experiment/nlp-v3) and maintain separate repos or branches for R&D work.

Project Management (Jira, Linear)

Create dedicated R&D projects or labels. Document technical uncertainty in ticket descriptions. Record experimental results in comments. Track R&D time separately from production tasks.

Technical Specs & Logbooks

Document the problem, hypothesis, methodology and results before and after each investigation. Use a wiki, Confluence or structured markdown. Entries must be dated and contemporaneous.

Frequently Asked Questions

Does building a SaaS product qualify for the R&D Tax Incentive?

Building a SaaS product can partially qualify. The portions of development that involve genuine technical uncertainty – novel algorithms, experimental architectures, performance challenges beyond known solutions – are eligible. Standard product development work such as building user interfaces, integrating payment gateways or implementing authentication using known methods does not qualify. Most SaaS companies find that 30-60% of their development work involves qualifying R&D.

Can agile development methodology be used for R&D claims?

Yes. The R&D Tax Incentive does not prescribe a specific development methodology. Agile, Scrum, Kanban and other iterative approaches are compatible with R&D claims. However, you must be able to identify which sprints, stories or tasks relate to R&D activities and which relate to routine development. The iterative nature of agile can actually help demonstrate systematic investigation – each sprint represents a cycle of hypothesis, experiment and evaluation.

Are contractor and freelancer costs eligible for the R&D Tax Incentive?

Australian-based contractor costs are eligible if the contractor performs work on registered R&D activities. The costs must be at arm’s length and you must be able to demonstrate that the contractor’s work relates to core or supporting R&D activities. Overseas contractor costs require a separate Overseas Finding approval from AusIndustry before they can be claimed.

Can we claim R&D for developing internal tools?

Yes, if the internal tool development involves genuine technical uncertainty. For example, building a novel data pipeline that must handle data volumes or processing requirements beyond what existing tools can manage may qualify. Building an internal dashboard using standard BI tools would not qualify. The key test remains technical uncertainty, regardless of whether the output is customer-facing or internal.

How does the R&D Tax Incentive interact with software grants?

The R&D Tax Incentive can be claimed alongside other government grants. However, you must reduce your R&D expenditure by the amount of any government grant received for the same activities. Double-dipping – claiming both the R&D offset and a grant for the same expenditure – is not permitted. Each program has different eligibility criteria, so it is possible to claim R&D on development work and an EMDG on marketing the resulting product.

What records does AusIndustry expect for a software R&D claim?

AusIndustry expects contemporaneous documentation that demonstrates technical uncertainty existed, a systematic investigation was conducted and new knowledge was generated. For software companies, this includes git commit histories, technical design documents, experiment logs, project management records (Jira, Linear), architecture decision records, benchmark results and meeting notes discussing technical challenges. The documentation should be created during the R&D, not retrospectively assembled at claim time.

R&D Tax Incentive by Industry

Get Expert R&D Tax Advice for Your Software Company

Identifying eligible software R&D, structuring claims correctly and maintaining defensible documentation requires specialist expertise at the intersection of software engineering and tax law. Prime Innovation works with software companies across Australia.

Contact Prime Innovation

R&D Tax Incentive Overview | R&D Tax Incentive Calculator

R&D Tax Incentive for Software Companies