The recent Camalic Pty Ltd vs Innovation and Science Australia (ISA) case at the Administration Appeals Tribunal (AAT) is the third court decision involving software development activities claimed under the R&D Tax Incentive during the past 18 months. The other three decisions on R&D handed down by the courts during this period were in the mining industry, with only one of these six cases, the Federal Court Moreton Resources Ltd vs ISA, found in favour of the taxpayer. Combined with the findings of the Australian Small Business and Family Enterprise Ombudsman’s (ASBFEO) report, the ISA could be perceived as making unfavourable determinations on software development in comparison to other industries.
We examine this latest case and consider whether it will provide further legal precedence for AusIndustry and/or claimants of software development activities under the R&D Tax Incentive.
Camalic Pty Ltd registered the core activity of “Development of an algorithm to predict shareholder value increases” along with three supporting activities:
- Acquisition of a supporting framework and database for the predictive tool
- Literature and technology review
- Project planning with AusIndustry.
The objective of the project, spanning 1 August 2013 to 31 December 2016, was to develop a predictive software tool utilising a significantly large number of multiple variables to predict an increase in value based upon the optimal combination of variables.
The core activity hypothesis was “that through utilising a Bayesian statistical approach, a predictive tool algorithm can be developed to accurately predict the increase in shareholder value that will result from improvements to various multi-dimension variables aspects of board governance.”
Based on the evidence, including statements from Camalic Pty Ltd, the AAT concluded the applicant never intended to build a bespoke machine learning algorithm. Instead, the objective of the project was deemed “to train an existing machine learning algorithm, using the functionality inherent in the algorithm, to adapt itself to predict increases in shareholder value from board governance variables”. The AAT was not convinced the claimed activities were intended to generate a bespoke algorithm or new, previously unknowable, functionality in an existing algorithm. Accordingly, the AAT affirmed it was not a core R&D activity for the purposes of the R&D Tax Incentive scheme.
The AAT further added that the collection and analysis of board governance data for the purpose of determining relationships between board governance variables and increases in shareholder value, came within the “management studies” exclusion and was, therefore excluded from being a core R&D activity. Finally, as there were no core R&D activities deemed to have been undertaken or planned by the applicant, none of the claimed supporting activities were considered supporting R&D activities for the purposes of the R&D Tax Incentive Scheme.
While we do not know all the details of the case, and many of the findings from it are specific to Camalic Pty Ltd’s R&D claim, it has significance to companies undertaking projects involving machine learning. Data scientists may consider the approach to solving problems using machine learning as experimental and likely falling within the ambit of eligibility under the R&D Tax Incentive program. However, we should now consider how the AAT came to it’s decision based upon the key requirements of what constitutes an eligible core R&D activity.
A key requirement of a core R&D activity is that it must generate new knowledge. In this case, both the AAT and ISA were of the view that new knowledge should be in the form of a new or improved machine learning algorithm. However, Camalic Pty Ltd did not claim to have intended to create a new machine-learning algorithm, rather the new knowledge was in the outcome of the machine learning. The view that if existing knowledge is used in an experiment it cannot qualify as eligible R&D is concerning, because this would significantly restrict the scope of activities that could qualify for the incentive. For example, trials of existing drugs for the treatment of COVID-19 would not be eligible. Indeed, in the Moreton Resources Ltd vs ISA case, the Federal Court found there was no basis in the legislation for excluding activities with the purpose of generating new knowledge with respect to the application of existing technology at a new site. The Federal Court also drew attention to the purpose of the incentive, which is to generate new knowledge or information in either a general or applied form. Unfortunately, in this case the AAT did not consider applying the new knowledge test in this way.
Another key requirement of core R&D activities is that the outcome of the experimental activities could not be determined in advance, it can only be determined by applying a systematic progression of work proceeding from a hypothesis to experiment and drawing of logical conclusions. The AAT agreed with AusIndustry’s view that Camalic Pty Ltd’s application and supporting material focussed on the uncertainty and knowledge gaps in the data being analysed rather than the uncertainty, knowledge gaps and experimentation to be undertaken in developing the functionality of the algorithms that would analyse the data. Despite seeming to accept the activities, with respect to deciding upon the right variables, involved experimentation, the AAT focused on whether experimentation was required to develop a suitable algorithm to solve the problem. The AAT accepted the testimony of the ISA expert witness, who was of the view the task was not out of the ordinary or difficult given the right data set and that machine learning algorithms were readily available and could be used to solve the problem. Again this view seemed to focus on creating the algorithm rather than creating the right data set. The requirement is whether the outcome of the experimental activities could not be known in advance, not on whether the task is an ordinary one. The AAT seemed to miss that the ISA expert actually agreed the outcome could not precisely be known before conducting analysis of the data.
Ultimately, the Applications and the way the project was scoped did not help the applicant and perhaps led ISA and the AAT to focus on determining whether a new or improved algorithm had or would be created. One of the initial concerns raised by AusIndustry had been the lack of information about the experimentation in the Applications. The core activity descriptions in Camalic Pty Ltd’s R&D Application remained largely the same across each subsequent claim. Whilst Camalic Pty Ltd later provided evidence with regards to deciding upon the right variables in the data, little detail appeared to have been provided to substantiate the claim that such research had ever been conducted, how the testing would be conducted, or how success would be measured.
Although this decision makes it clear that AusIndustry, on behalf of ISA, will have concerns about eligibility in projects involving data science and machine learning, the decision and approach affirmed by the AAT should not be a cause for concern for most software development claimants. We have seen several positive findings by AusIndustry for projects involving data science and machine learning where claimants have substantiated that the process of getting to the outcome, rather than just the outcome, can represent new knowledge. This case highlights the importance of contemporaneous evidence that can substantiate the technical unknowns by using technology and how these uncertainties can only be resolved through a systematic experimental approach.
Whilst it is disappointing there had been little to no consideration of the Federal Court’s decision in the Moreton Resources case on the use of existing technology in eligible core activities, a positive from this case is that the development of a bespoke algorithm or new functionality from an existing algorithm could potentially satisfy the core R&D activity requirements. We are aware of an upcoming appeal to the Federal Court that may provide some further clarity on what constitutes eligible software R&D. In the interim, most software claimants would prefer AusIndustry and the AAT to apply the approach taken by the Federal Court in the Moreton Resources case.
If you would like further information on the R&D Tax Incentive, please contact a BDO R&D Tax Adviser today.
This article was originally published in the July edition of the AIIA Connector Magazine.