Hypothesis Driven Software Development

Posted on by Admin

While hypothesis-based testing is natural for most scientific work, it doesn’t come as easily nor is it as widely used in the software development space. One of the reasons may be the rise of Agile and the “fix it faster” culture. There is little time to document the systematic thought processes that takes place in programmers’ heads. However, it is exactly this journey in the thinking process that moves from question or idea to a proposed solution and implementation, then testing and results analysis that AusIndustry want to see as evidence for the R&D Tax Incentive.

Applying the scientific process to software development

It is not that difficult to adopt a scientific thinking process to traditional and disruptive software-based products, processes or services (the three generic avenues through which R&D can be claimed). For organisations that use a more traditional, Waterfall methodology, the scientific methodology fits in neatly in that the higher emphasis on early documentation can be used to help record various hypotheses about functionality, performance or other measurable attributes related to the project in question.

For organisations using an Agile methodology, a move towards the Lean model of Build-Measure-Learn will ensure there is enough scientific rigour in the methodology adopted to be able to demonstrate that the project met AusIndustry’s eligibility criteria.*

Creating hypotheses and incorporating IT projects

A Hypothesis-Driven-Development approach to IT-based projects not only helps in meeting the eligibility criteria for claiming the R&D Tax Incentive, it ensures that companies gain validation of ideas relatively quickly.

For startups, the questions about whether the functionality offered will meet particular metrics will form broad, overarching hypotheses. Will my product solve a current problem or do it better than existing products? Will I be able to scale as quickly as I can if I adopt technology x? These questions and more can be answered through several rounds of prototyping using the Lean model. To claim the R&D tax incentive (provided that the offering is novel on a worldwide basis), these questions can (and should) be drilled down further into specific technical questions e.g. “The <<novel, newly created function>> will calculate < > within x seconds.” Metrics ranging from memory utilisation, accuracy in results, latency and a range of other factors are often tested and will form good measurable outcomes for the questions asked in the hypothesis statements.

In a more established organisation, there may be templates and methodologies that are followed in developing new products, processes or services. With minor tweaks to these processes, you can adopt a style of thinking which will identify the most risky assumptions, conduct tests on them and form conclusions.

Gathering test data and forming conclusions

Keeping supporting documentation and evidence of tests conducted is an essential criterion for claiming the R&D tax incentive. Fortunately, the hypothesis-driven-development (HDD) approach lends itself well to capturing test results and reaching conclusions. Agile companies often use a concept of user stories to validate requirements and planned functionality. HDD asks that you develop acceptance criteria for each story. To conduct strong qualitative and quantitative result analyses and therefore validate assumptions (hypotheses) with greater certainty, statistical hypothesis testing is recommended.

If your business is undertaking R&D, call Access RnD Tax Solutions for an obligation-free discussion on the eligibility of your projects and for support in setting up systems to capture contemporaneous evidence in the above format. Call us today on 1800 RnD TAX(1800 763 829).

*Note that the entire project will not usually be eligible to meet the R&D tax incentive criteria. It will come down to the novel products, processes or services developed to which there was no answer to in the public domain and involved significant technical risk.

Some helpful tools and techniques

 

Posted in R&D