The tax challenge for R&D tax credit claims for social media projects is defining what the technology “baseline” is and then demonstrating and evidencing the “advance” in resolving scientific and technological uncertainty. HMRC of the UK and tax authorities in other jurisdictions too, often do not understand the key scientific parameters within the social media space and therefore can give strong “push-back” on innovation claims. Sadly, the “default” approach taken by many tax administrations is that the scope for scientific innovation within the social media space is limited as technology is largely based on well-established “legacy” systems and any changes are therefore not true “advancement” but the “re-arrangement” of existing technology. Nothing could be further from the truth as the developments in the social media space are a “fluid” not a “static” process as user interests and demands are constantly evolving, which necessiates continual scientific and technological innovation.
In defining the project’s scientific “baseline”, the starting approach might be to explore the particular Artificial Intelligence (AI) or Machine Learning (ML) algorithms that might apply to the project based on the “use case” and the “type of problem” the algorithm is trying to solve. For example if the particular “use case” / “problem solving” involves social media Recommendation Engines based on Collaborative or Content-based filtering techniques, the AI platform might employ a “Wide and Deep” algorithm which is often used to address large scale classification and regression problems such as those encountered in recommender systems, search and ranking problems.
The ML model employed might be a Tensor Flow Estimator. This type of model combines a linear model that learns and “memorizes” a wide range of rules with a deep neural network that “generalizes” the rules and applies them correctly to similar features in new, unseen data. The “baseline” might be using so-called “legacy” algorithms such as SVD++ a form of “matrix factorisation” or Restricted Boltzman Machines (RBM) which are neural networks adapted to work in collaborative filtering for Recommendation engines.
The key is demonstrating how these algorithms have been modified and combined in linear or other ways and applied to the project’s particular datasets to produce higher accuracy estimates that will demonstrate to tax authorities that there has indeed been true “innovation” .