Why your whole project is not your R&D claim

Most business owners and founders describe an ambitious, multi-phase commercial project when they first think about the R&D Tax Incentive. They might say, “We are building a new software platform,” or “We are developing a better manufacturing process.” The instinct is to frame the entire project as the R&D claim. That instinct is understandable, but it is also the single largest reason ATO and AusIndustry scrutiny focuses on activity identification and cost apportionment.

This article explains why your whole project is almost never your R&D claim and provides a step-by-step guide to isolating the eligible core and supporting activities from the broader commercial effort. It is general information only, not tax, financial, or legal advice. Always confirm your specific circumstances with a registered tax agent who understands the incentive rules. For a plain-English introduction to the incentive, refer to our What is the R&D Tax Incentive? A plain-English guide for Australian businesses.

GrantsMAX prepares an evidence-backed application pack from your accounting data (Xero, MYOB, QuickBooks, Microsoft 365, and Google Workspace, read-only); a registered tax agent reviews, refines, and lodges, and your business owns the claim. We never lodge, guarantee outcomes, or file on your behalf.

Prerequisites for isolating eligible activities

Before you can split a project into eligible and non-eligible components, you need a few things in place:

• A working understanding of the R&D Tax Incentive eligibility rules as set out by the ATO and AusIndustry, particularly the definitions of core and supporting R&D activities.
• A clear, documented scope for the broad business project, what you set out to build, achieve, or improve.
• Access to your financial and operational records (general ledger, payroll, timesheets, project management logs). GrantsMAX connects to these to draft narratives and cost structures, but the raw data must exist.
• A willingness to examine your project through a lens of “scientific or technological uncertainty” rather than just commercial success.

If you haven’t yet assessed eligibility, the Eligibility Assessment & Risk Flags feature within GrantsMAX can give you an initial read based on your own accounting data, but it does not replace a formal determination by a tax agent or AusIndustry registration. It flags where a reviewer would look hardest, so you know where to concentrate your evidence.

Pro tip: Engage your registered tax agent early. They can advise on the eligibility of activities before you scope the claim, which saves rework later.

Step 1: Define the overarching business project

The first step is to clearly describe the commercial initiative you undertook. This is the “project” that might include research, development, production, marketing, and everything else. For example: “Develop and launch a cloud-based analytics dashboard that predicts customer churn for mid-market retailers.”

Notice the project includes designing algorithms, coding the platform, testing performance, creating training materials, running sales trials, and deploying the product. Only some of those activities will meet the R&D eligibility test. The project is the container; the R&D claim is the subset. As the ATO explains, not all expenditure on a project with R&D elements qualifies. You must identify the specific activities that directly contribute to resolving technical uncertainty.

Warning: Do not use the project title or high-level description as the activity description in your claim. The ATO and AusIndustry require claim narratives that are specific, technical, and tied to the scientific or technological advance you sought.

Step 2: Break the project down into all tasks, phases, and activities

Map every discrete task you performed during the project. Use your project plans, developer logs, sprint boards, or timesheets to list tasks. For a software project, this might include:

• Literature review and technology scoping
• Requirements gathering and specification writing
• Core algorithm research and experimentation
• Prototype development and iterative testing
• UI/UX design and front-end coding
• Integration with legacy systems
• Beta testing with customers
• Documentation and training material creation
• Marketing and sales activities
• Post-launch maintenance and bug fixes

This exhaustive list is the raw material from which you will select eligible R&D activities. It ensures nothing is omitted and that you can later demonstrate why a task did not qualify.

Step 3: Identify the core R&D activities

At the heart of the Australian R&D Tax Incentive are “core R&D activities.” The ATO and AusIndustry define these as experimental activities whose outcome cannot be known or determined in advance on the basis of current knowledge, information, or experience, and which are conducted for the purpose of generating new knowledge. They must be systematic, investigative, and experimental, and they must seek to resolve a scientific or technological uncertainty.

The UK’s HMRC guidance uses a similar test: the work must seek an advance in science or technology and resolve scientific or technological uncertainty. Both frameworks ultimately ask the same question: did you run experiments to overcome a technical obstacle for which the solution was not readily deducible by a competent professional in the field?

In the analytics platform example, the core R&D activity might be “experimental development of a machine learning algorithm to predict churn from incomplete and noisy transaction data.” Designing the experiment, building prototype models, testing hypotheses about feature engineering, and evaluating accuracy against uncertainty are all potentially core activities. In contrast, writing a standard REST API to serve the dashboard probably is not.

To decide whether an activity is core, ask:

• Could a competent professional in the field have worked out the outcome using publicly available knowledge or established techniques? If yes, it is unlikely to be a core R&D activity.
• Did you conduct experiments to test multiple possible solutions because none were known or obvious? If yes, you likely have a core R&D activity.
• Did you document the hypotheses, methods, and results? Without contemporaneous records, you may struggle to evidence the activity.

Step 4: Identify the supporting R&D activities

Supporting R&D activities are activities that are directly related to core R&D activities. They do not themselves need to meet the experimental test, but they must be undertaken for the dominant purpose of supporting the core R&D. Examples from ATO guidance include:

• Building and testing prototypes used in core experiments
• Data collection and cleaning when directly feeding into an experimental process
• Scientific or technical training required to carry out the core activity
• Activities to protect the knowledge generated (e.g., patent applications) when directly linked

In our software project, supporting activities could include:

• Developing a data pipeline to ingest and clean retailer transaction logs for the algorithmic experiments
• Constructing a sandbox environment to run controlled A/B tests on model variants
• Drafting a provisional patent application for the novel churn-prediction method

A critical condition is that the supporting activity must produce goods or services that are only used for the core R&D, or, in some cases, that the activity itself is a necessary input. The dominant purpose test is strict. For instance, building a customer-ready interface is usually not a supporting activity because its dominant purpose is commercial, not experimental.

Pro tip: When documenting supporting activities, tie them explicitly to a core activity in your records. Write something like, “This data cleansing script was written solely to prepare training data for the uncertainty-resolution experiment on 12 July,” and keep the code commit log.

Step 5: Isolate the non-eligible activities

Now pull out everything that is not core and not supporting. These activities cannot be included in the R&D claim, even if they are critical to the project’s success. Typical non-eligible activities include:

• Commercial production, manufacturing, or deployment of a product or service that has already been proven
• Routine software development using established techniques
• Market research, sales, and customer support
• Administration, compliance, and general management
• Training end users or creating user documentation
• Post-launch maintenance and bug fixes that do not involve new technical uncertainty

In the analytics platform, the front-end development that follows a well-known framework pattern, the marketing campaign, and the customer success training are all non-eligible. Their costs stay out of the R&D claim.

Step 6: Apportion costs strictly to eligible activities

Once you have the activity map, attach costs only to the eligible core and supporting activities. The ATO allows you to claim the R&D tax offset on expenditure that is directly incurred in carrying on those activities. This usually includes:

• Salaries and wages of employees who performed the eligible activities (on a time-spent basis)
• Materials consumed during experiments
• Depreciation of plant and equipment used directly in the R&D (apportioned)
• Certain contract expenditure
• Overhead costs that can be reliably attributed

You cannot claim the whole development team’s salaries for a year. You must apportion. For example, if a software engineer spent 60 per cent of their time on the experimental algorithm and 40 per cent on general front-end work, only the 60 per cent of their salary attributable to the core and supporting activities may be eligible.

Detailed time-tracking systems, such as those integrated into practice management or project tools, become essential evidence. The Audit-Ready Evidence Trail that GrantsMAX builds indexes emails, invoices, and timesheets and ties each cost line in the pack to its source. That linkage is what the ATO expects if a claim is reviewed.

International experience underscores this. In the United States, the IRS’s Form 6765 and accompanying instructions now require taxpayers to identify the specific business components to which research costs relate. Thomson Reuters reports that the IRS is “continuing to fight R&D tax credit abuse” by demanding claim-specific documentation that ties costs to activities. The Tax Adviser notes a new era of disclosure where vague project descriptions no longer suffice. While these are US rules, the trend is mirrored in Australia’s ATO review programs. Contemporaneous, activity-level evidence is not optional.

Warning: Overclaiming by treating the entire project as R&D or by including ineligible expenditure is the fastest way to draw ATO scrutiny, penalties, and reputational damage. A registered tax agent will refuse to lodge a claim that lacks proper apportionment.

A worked example: AI-based analytics platform

Let’s walk through the method with a fictional mid-market software company, Retainly Pty Ltd. Retainly set out to build a predictive analytics platform for retailers. The CEO originally thought, “The whole platform is R&D because it’s innovative.” Here’s how a properly isolated claim would look.

The project

“Design, build, and commercially launch a cloud-hosted platform that ingests retailer transaction data and predicts customer churn with >85 per cent accuracy.”

Full activity breakdown

1. Market and technology scoping
2. Literature review on churn prediction models
3. Design of a novel ensemble model architecture
4. Iterative experimentation with candidate algorithms and hyperparameters
5. Data collection, cleaning, and feature engineering
6. Development of a proprietary feature store for model training
7. Integration testing with a sample retailer’s legacy POS system
8. Building a customer-facing dashboard UI using a standard framework
9. Beta release to three test retailers
10. Sales and marketing collateral preparation
11. Staff training on the new platform
12. Ongoing support and feature updates post-launch

Eligibility analysis

• Activities 2, 3, 4, and 5: Core R&D. They involve systematic experimentation to resolve the technical uncertainty of predicting churn from noisy, sparse data. Literature review guides the experiments, model design and testing are the experimental activities, and data preparation is part of the experiment.
• Activity 6: Supporting, if the feature store is built solely for the core experiments and not later sold as a separate product.
• Activity 7: Supporting, because integration testing with real-world data is a necessary step to validate the model’s performance under experimental conditions.
• Activity 1: Non-eligible (market scoping).
• Activity 8: Non-eligible (routine development).
• Activity 9: Non-eligible (beta testing for commercial validation, not systematic experimentation).
• Activities 10 to 12: Non-eligible (commercial and maintenance).

Cost apportionment

Retainly spent $800,000 on the project. After allocating staff time, contractor costs, and materials to the eligible activities, the total qualifying R&D expenditure came to $320,000. The balance was spent on commercialisation and routine development. The claim preparation included timesheet extracts, commit logs, experiment notebooks, and a narrative that described each core activity in technical detail.

The result? A defensible claim that stands up to review, not a mass of unsupported project costs.

Pro tip: Use a consistent taxonomy across your project management tool, timesheets, and accounting software. Tag tasks as “core-R&D,” “supporting-R&D,” or “commercial” from the outset. That makes end-of-year apportionment far simpler and creates a contemporaneous record.

How GrantsMAX fits into the isolation process

GrantsMAX does not decide which activities are eligible, that is the role of your registered tax agent. But it dramatically reduces the manual work of drafting the claim pack. Here’s how:

• The platform reads your Xero, MYOB, or QuickBooks data and automatically pulls the ledger categories and contacts that are likely to contain R&D expenditure.
• Our AI Application Pack Drafting feature then generates draft activity narratives and a cost structure, mapping costs to specific activities based on the accounting data and your inputs. It produces a complete, evidence-backed pack, drafted in hours, not weeks.
• The pack is handed to your registered tax agent through the Accountant Review & Lodge Workflow, where they can review, refine, and lodge the claim. The business owns the claim, and the accountant remains in control at every step.
• Throughout the process, the Eligibility Assessment & Risk Flags engine flags areas a reviewer would look at, prompting you to strengthen evidence before the accountant signs off.
• The final output ties each cost line to source documents, creating an Audit-Ready Evidence Trail your accountant can stand behind.

If this is your first R&D claim, it can feel intimidating. GrantsMAX for first-time claimants shows how we turn your business data into a pack your accountant can review, so you’re not starting from a blank page. For software and technology companies, GrantsMAX for technology companies explains how we draft the technical narratives and substantiation that make the difference between a rejected claim and a robust one.

Compared to traditional consultants who may charge 10 to 20 per cent of your claim as a success fee, GrantsMAX vs R&D consultants highlights a faster, lower-cost approach that keeps the process in your hands, with your accountant at the helm. And if you already have an in-house finance team, GrantsMAX vs In-house finance team shows how automation cuts the manual weeks down to hours.

Common mistakes when isolating activities

Even experienced claimants slip up. Here are frequent errors and how to avoid them:

1. Treating the whole project as core R&D. As this article has stressed, the project is the business case, not the R&D activity. Apply the “uncertainty” test relentlessly.
2. Confusing innovation with technical uncertainty. A product can be new to the market without breaking new technical ground. The ATO requires a systematic progression through scientific or technological uncertainty, not just a clever combination of known technologies.
3. Failing to document the failed paths. Experiments that don’t work are often the strongest evidence of uncertainty. UK guides such as SeedLegals remind claimants to describe “what you attempted and why it didn’t work.” Keep lab notebooks, decision logs, and version-control histories.
4. Claiming costs that would have been incurred anyway. Rent, general-purpose software licences, and baseline employee salaries are not directly incurred on R&D unless you can apportion them. Overheads need a defensible allocation methodology.
5. Skipping the AusIndustry registration. You must register your R&D activities with AusIndustry within 10 months of the end of the income year. Failure to register invalidates the claim, no matter how good your activity isolation is.

Warning: Never let the size of the potential offset tempt you to claim marginal activities. The ATO’s compliance program uses data analytics to spot patterns, and advisers face Tax Practitioners Board obligations. Your registered tax agent will want to see a clear, documented basis for every activity you include.

Record-keeping that supports the isolation

Good isolation starts before the claim year ends. ForrestBrown, a specialist UK firm, offers a practical guide that translates well to Australian requirements: you need records that show “what the advance was, what the scientific or technological uncertainties were, and why the expenditure was incurred.” In the Australian context, extend that to include:

• Detailed timesheets that tag time against specific activities (core, supporting, commercial)
• Technical logs or notebooks that record hypotheses, experiments, observations, and conclusions
• Emails and meeting notes that discuss technical challenges and decision points
• Prototype versions, data sets, and test results with date stamps
• A clear project plan that separates R&D phases from commercial phases

The Audit-Ready Evidence Trail in GrantsMAX indexes these items and links them to the cost lines, so the pack your accountant reviews is already structured for review-readiness. But the underlying records must exist and be complete. AI can’t invent timesheets or lab notes.

Key takeaways

• Your whole project is not your R&D claim. Only the activities that involve systematic experimentation to resolve a scientific or technological uncertainty, and the activities directly supporting them, can be included.
• Isolate activities by first defining the broad project, then breaking it down into discrete tasks, classifying each as core R&D, supporting R&D, or non-eligible, and finally apportioning costs.
• The example of Retainly’s analytics platform illustrates how a seemingly innovative project might have a large non-eligible portion.
• International guidance from the UK’s HMRC, the IRS, and specialist firms reinforces that contemporaneous, activity-specific documentation is the global benchmark.
• GrantsMAX accelerates the preparation of an evidence-backed pack, but it does not make eligibility decisions or lodge claims. A registered tax agent does that, ensuring the claim meets the Tax Practitioners Board standards.

If you’re a founder, CFO, or accountant supporting clients, and you’re ready to move beyond the “whole project” approach, consider GrantsMAX for founders and CFOs or GrantsMAX for accounting and bookkeeping firms to see how we streamline the pack preparation. You can also explore the full Grant & R&D Discovery and Matching engine, which continuously scans for state innovation grants and the EMDG export grant in addition to the R&D Tax Incentive.

Join the GrantsMAX waitlist today to be among the first to access a smarter way to claim. Your accountant, and your audit trail, will thank you.