This article is for people inside Australian companies who are trying to work out whether a particular piece of work qualifies for the R&D Tax Incentive. Whether the company can claim at all is a separate question, so the focus here is the work itself. It explains what counts as R&D in plain English, including the tests that apply and the questions that resolve most cases. It also covers the work that never qualifies.
Legislation sets out the definition in technical language, which is why the same words get read two opposite ways. Some read it as “anything new to us” and claim far too widely, while others read it as “science” and claim nothing at all.
The one idea underneath the whole test
Whether the work was hard, expensive, innovative or commercially important has no bearing on eligibility. What matters is whether a competent professional in the field could have worked out the outcome before you started, using current knowledge, information or experience.
If somebody could have looked it up, asked a specialist, read the manual or reasoned it out from what is already known, the work is ineligible however much it cost you. If you needed to run the experiment to reach the answer, you are in the right territory.
Buying and installing a new machine sits outside the definition, while modifying that machine to run a material it was never designed for can sit inside it. Building a website from known components does not qualify, though solving a performance problem that the available techniques could not solve might.
Core activities, in plain English
The ATO’s guidance on eligibility for the R&D Tax Incentive sets out the characteristics of core R&D activities. Current knowledge, information or experience must not already determine the outcome. That uncertainty is measured against the state of knowledge in the field, not simply what your own team happens to know.
A systematic progression of work must then determine the outcome using principles of established science. That progression moves from hypothesis to experiment, observation and evaluation, leading to logical conclusions. In practice this means you had a stated expectation, tested it deliberately and recorded what happened. You then drew a conclusion that changed what you did next. Trial and error with no hypothesis and no record does not clear this bar, even when it eventually works.
The activities must aim to generate new knowledge, including knowledge about new or improved materials, products, devices, processes or services. Commercial motivation is fine and expected. The purpose test asks what the activity was for, and an activity aimed squarely at making your process work better can still generate new knowledge about that process. The legislation also contains a list of activities specifically excluded from being core R&D activities, so check that list directly.
Supporting activities, and where people over-reach
Supporting R&D activities are those directly related to core R&D activities. This includes work needed to run the experiment, such as building a test rig, preparing samples or analysing the results so you can draw conclusions.
Activities that produce goods or services, or are directly related to producing them, face a tighter dominant purpose test. The same applies to activities on the excluded list. They count only where their dominant purpose was to support core R&D activities.
Where a production run’s dominant purpose was making product for sale, useful experimental data generated along the way does not make the run eligible. Manufacturers encounter this often because the experiment and the production line are frequently the same equipment on the same day. It is a recurring theme in our manufacturing R&D work.
Two questions that resolve most cases
The first is whether, before you started, a competent person in your field could have told you the answer. An honest yes closes the matter. An honest “they would have had a view, and it turned out to be wrong for reasons we could not have anticipated” can be the beginning of a real claim.
The second is what you wrote down at the time. A contemporaneous record should show the hypothesis, test conditions, results and the decision that followed, with dates. Documentation does not make ineligible work eligible, and its absence can make eligible work unclaimable. That is one of the ways businesses lose money on this program.
That is why it helps to set up the record keeping at the start of the income year, because by the following April the memory of which fortnight the failed trials ran is already gone.
Work that does not count, however hard it was
Buying, installing and commissioning equipment to the supplier’s specification falls outside the definition. Routine software development using established frameworks and patterns also sits outside it, however complex the finished product.
Market research, market testing and customer trials aimed at commercial questions do not qualify. The same applies to adapting a known solution to your own circumstances where the adaptation follows from what is already known. Quality control and routine testing to confirm that things work as expected do not count, and neither does management, sales or the ordinary work of running the business.
Commercial importance does not determine R&D eligibility. Plenty of this work may be more commercially important than the R&D sitting next to it, yet it falls outside what this program funds.
The same project can be partly eligible
Businesses often assess eligibility at the level of the project, although the legislation applies the tests to individual activities. A twelve month product development project is unlikely to be eligible from end to end. The work at the front, where the team is scoping and specifying, is not experimental.
A genuine technical problem may then emerge and the team may work through it experimentally. That period may well contain core R&D. You need to assess the supporting work around it separately, while packaging, documentation and commercial release are ordinary business activity.
Claiming the whole twelve months can create real exposure where only part of the project is eligible. Ruling out the whole project can also mean losing a legitimate claim on the eligible period in the middle.
The right unit of analysis is the activity, with the time apportioned honestly between them. That apportionment requires judgement supported by evidence, which is exactly why contemporaneous records matter so much.
Failure counts, success is not required
The program does not require the experiment to work. An experiment that resolves a genuine uncertainty by demonstrating that an approach cannot work has generated new knowledge. Abandoning the project afterwards does not undo that. Businesses sometimes leave eligible work out of a claim because it “did not go anywhere”, even though the abandoned work can provide clear evidence of genuine technical uncertainty.
Working out what counts as R&D in your business
The practical exercise is to list every project that consumed meaningful engineering, development or technical time in the year, then run each one through the two questions above. Some projects will be straightforward, while others will depend on facts that have not yet been written down. Those unresolved projects are where careful review matters most.
We do this work across sectors through our R&D Tax Incentive advisory practice, including alongside the ordinary accounting we do for manufacturing and distribution businesses. The answer often depends on operational detail more than on tax technique.
If you want a straight view on what counts as R&D in your business, including a clear no on the projects that do not qualify, book a call at pp.tax/contact/ and bring your project list.