It’s always been my view when looking at projects in machine learning that, at their heart, the real value lay in the data (the big data component) and that the math behind them was almost secondary. ‘It was always better to have good data and average algos than the other way around’ the saying went.