Hypothesis: definition for R&D Tax Incentive claims

Hypothesis: the testable proposition at the start of every core activity

In the R&D Tax Incentive, a hypothesis is the specific, testable proposition a core R&D activity sets out to confirm or disprove, recorded before experiments start. It names the result sought and why it may or may not be achievable, and it is the first step of the systematic progression section 355-25(1) requires.

Reviewed by George WalchFacts last verified 25 October 2026

Example

For example, an illustrative proptech company first wrote "build faster, smarter property search", which is a goal. Its usable hypothesis was that a hybrid keyword and embedding ranker would lift nDCG@10 by at least 8% over keyword ranking alone on its 40,000-query search log, while keeping p95 latency under 120 ms; disproved if no weighting met both limits.

The Act requires work that "proceeds from hypothesis to experiment". AusIndustry asks you to develop the hypothesis before the core activity starts, stating "what result you aim to achieve" and why that result may or may not be achievable.

What makes a hypothesis hold up

A hypothesis that survives review is:

specific, naming the mechanism, method or parameter under test;
testable, by an experiment designed so that its result can confirm or disprove it;
falsifiable, with the result that would disprove it written down;
measurable, with a threshold or a pre-specified endpoint;
grounded in the prior-art search, so it reflects what the competent professional could not predict;
technical. "Feasible", "cost-effective" and "scalable" state no unknown on their own. If scale or cost is the real constraint, name it as a measure, such as throughput at a stated load or compute cost per inference, not as a business result.

A hypothesis records the unknown; it does not create one. If a competent professional could have predicted the result, a well-written hypothesis does not make the work core R&D.

Where hypotheses have failed

Tribunals read this field closely. In Royal Wins [2020] AATA 4320 the Tribunal said a hypothesis is a proposition tested by experiment, not an objective, and found no record that one was developed. In Absolute Vision Technologies [2022] AATA 2319 the hypotheses were project goals. In Active Sports Management [2023] AATA 4078, upheld on appeal in [2024] FCA 1346, nothing showed how hypotheses were formed. The common thread is a goal written as a hypothesis, often after the fact.

How it is used in a claim

The registration form gives the hypothesis its own field for each core activity, and the experiment and conclusion fields must answer it. If an earlier record described the hypothesis wrongly, a dated clarification can explain the error, but it is not itself a contemporaneous record. Worked hypotheses for AI and algorithm work are on novel algorithms and LLM fine-tuning. It is the first link in the systematic progression.

Frequently asked questions

Can the hypothesis change during the activity?

Yes, and it often should. A conclusion that disproves one hypothesis usually leads to the next. Record each new hypothesis, dated, before testing it. What does not work is rewriting the original after the results are known.

Does a hypothesis have to include a number?

Not always, but it needs a measurable test: a quantitative threshold or a pre-specified clinical or behavioural endpoint. Without one there is no way to say whether the experiment confirmed or disproved it.

The R&D Tax Incentive is a self-assessment program. This page is general information, not tax, legal or financial advice; eligibility depends on your circumstances and you should seek independent advice for them.