Moving a process from the bench to production scale is expensive, slow and often fails, which makes it feel like research. Expense and difficulty are not the test. The question is whether the outcome at the larger scale could be known in advance by a competent process engineer working from the smaller-scale data and published knowledge.
The test: name the effect that does not scale
"Never made at this volume" is not an uncertainty, and AusIndustry says stating that a scale-up is new is not enough. A scale-up activity can be core only where the company can point to a physical reason the smaller scale cannot predict the larger one, and design runs to resolve it. Typical reasons:
- Heat transfer. Surface area falls relative to volume, so cooling and heating rates change.
- Mixing and mass transfer. Gradients appear that a small vessel never had.
- Kinetics and residence time. Competing reactions shift when hold times lengthen.
- Impurities and phase behaviour. Trace components build up in recycle streams, or a different crystal form or emulsion appears.
If the larger-scale behaviour could have been modelled or calculated from bench data, running it is engineering, not an experiment. Record the named effect, and the reason modelling could not settle it, before the first larger run. A scale-up plan that only lists target volumes and dates reads as a production schedule, however uncertain the team felt at the time. In Coal of Queensland the Full Federal Court also drew a line between data and outcome: that test numbers cannot be predicted precisely does not mean the outcome was unknown.
Production runs: define the experiment first
Many scale experiments run on production equipment, and the equipment does not decide the answer. Start by defining the core activity: the designed test conditions, the runs that apply them and the evaluation. That activity is judged on its own purpose of generating new knowledge, and a batch run only to test a hypothesis can be the experiment throughout, even on the production line.
The production work around it is a separate question. Where testing is added to a campaign run to make product, the core activity covers only the periods and parts of the run where test conditions were applied, and costs are apportioned to those windows. The rest of the campaign produces goods, so it can be supporting only if its dominant purpose was to support the experiment, which a commercial campaign rarely shows. Routine monitoring between tests is neither core nor supporting, even if its data later informs the experiment. AusIndustry adds that choosing the best settings at scale by trial and error among available options, rather than testing a hypothesis, is not core.
The core activity also ends. Once the process is predictable it is business as usual, and repeating it at another site or line is not a new core activity unless a new knowledge gap is stated. Process validation, tech transfer of a settled process and capacity increases fall on the far side of that line. Testing done to show a product meets a standard is routine testing, set out on materials testing.
Saleable batches and feedstock
Experimental runs often produce something usable. Where R&D output is supplied to others or applied to the company's own use, the ATO's feedstock rules add an amount to assessable income to recover the incentive part of the offset on the inputs and energy consumed. The adjustment is based on the lesser of the feedstock expenditure and the feedstock revenue, and it also applies when an affiliate or connected entity sells or uses the product. Track every experimental batch to its destination so the company's tax adviser can work it out.
Equipment built to run the experiments, and how its cost is claimed, is covered on pilot plants and prototypes and tooling. Trial-stage drug developers will find the surrounding rules on clinical-stage pharma.