There's a well-understood pattern in R&D Tax Incentive claims: software firms over-claim relative to eligibility, and heavy industry firms under-claim relative to it.

The software pattern attracts most of the regulatory attention. The heavy industry pattern leaves significant amounts of legitimate refund on the table every year, in sectors where the cash matters most.

Three sectors where we see this consistently are hydrogen and clean energy, AgriTech and environmental data, and advanced manufacturing.

Why these sectors under-claim

The shared characteristic is that the R&D happens inside operating businesses, not inside dedicated technology companies. A hydrogen retrofit specialist running pilot installations on diesel trucks isn't structured as an R&D company — it's a working business that happens to be doing genuinely experimental work as part of how it delivers. An agricultural sensor platform deployed across a working pastoral operation looks like operations from one angle and like systematic field trials from another.

This creates three obstacles to claiming:

The work doesn't get categorised as R&D.

Engineering teams describe their work in operational terms — "we ran the trial", "we adjusted the configuration", "we got the sensor stack working" — rather than in hypothesis-driven terms. The eligibility lens needs to be applied externally; it doesn't surface naturally from how the team talks about what it did.

Evidence is embedded in operational records, not R&D records.

Trial data sits in production databases. Engineering decisions sit in jira tickets or design review minutes. There's plenty of contemporaneous documentation, but it's not labelled or structured as R&D evidence. Pulling it together for a claim requires going through operational records and selecting what's relevant.

Apportionment is genuinely hard.

A field engineer who spends 60% of her time on commercial deployments and 40% on trial work needs that split to be derived from records that already exist, not estimated retrospectively. Without time tracking discipline, the apportionment defaults to a guess, which the ATO won't accept.

The hydrogen pattern

In hydrogen retrofit and electrification work, the most commonly missed category is the systems integration experimentation. Retrofitting a diesel powertrain to run on hydrogen involves repeated testing of how the modified system performs under load, in different operating conditions, against different fuel feed configurations.

The team running the trials is solving real technical uncertainty — is the modified injection geometry stable at altitude, does the storage system handle the thermal cycle, can the safety interlocks be calibrated tightly enough to meet certification requirements — and the answer is genuinely not known in advance.

This work is eligible R&D. It often doesn't get claimed because the team describes it as "commissioning" or "trials" or "engineering iteration" rather than as experimental testing of a hypothesis.

The AgriTech pattern

In precision agriculture and environmental data, the under-claiming usually centres on the early phases of platform development. The work to develop a novel sensor approach, to validate its performance across the variability of a working pastoral operation, to refine algorithms against field-collected data, has structural eligibility — there's hypothesis, uncertainty and systematic experimentation throughout.

The challenge is that the same team running the experimentation is also running commercial deployments. Time gets attributed by default to whichever activity invoices a customer, rather than to whichever activity meets the R&D test. Without explicit apportionment discipline, the R&D portion of the work is invisible in the financial records.

We've seen multi-year programmes structured around environmental data platforms where the apportionment work alone resulted in registered expenditure five times higher than what the company had previously been claiming on a self-prepared basis.

The manufacturing pattern

In advanced manufacturing and on-shore production, the missed category is process development. A company developing a novel manufacturing approach — a new heat treatment, a novel forming process, an unusual material combination — typically claims the obvious experimental work (the lab trials, the prototype builds) but misses the iterative process refinement that follows. Each production run that tests a process modification, each batch that validates an unknown, each adjustment that resolves a technical uncertainty, is potentially eligible.

The apportionment question here is even sharper than in AgriTech. A production engineer who spends a shift adjusting parameters to dial in a new process is doing R&D, but only if the underlying question genuinely wasn't answerable in advance. The structuring conversation has to surface where genuine experimentation happened, not just where engineering effort was expended.

What to look for in your own operations

If you operate in heavy industry, clean energy, AgriTech, manufacturing or any sector where R&D happens inside an operating business rather than inside a dedicated R&D function, the questions worth asking are:

  • Where in the past income year did the technical team genuinely not know the answer in advance?
  • How often did a trial or production run change direction based on what the previous one revealed?
  • Where are the design decisions documented, even informally?
  • Who could speak to the hypothesis behind a particular piece of work, and what records would they reference?
  • What proportion of the technical team's time is genuinely directed at resolving open technical questions, versus delivering known solutions?

The answers tend to surface eligible work that wasn't being recognised as eligible work. Translating that into a defensible claim is a structuring exercise — applying the R&D Tax Incentive's eligibility lens to operational records that were never designed with the claim in mind.

The structural payoff

For under-claiming sectors, getting the structuring right doesn't just affect this year's refund. It changes how the company captures evidence going forward. Time tracking conventions get adopted. Decision points get documented. The next year's claim is materially easier to build because the operational data is already in the right shape.

This is the work most R&D consultancies don't do well in heavy-industry sectors, because it requires understanding the operational reality of the business rather than just applying a tax framework. The eligibility test is the same in every sector. The work to surface where the eligibility actually lives is different in every sector.

That's where the under-claiming gets resolved.