At the latest EMAC 2016 conference, Michel Wedel gave a brilliant demonstration of how data collection methods have evolved over the last 100 years and how modelling changed. His graphical representation of (Big) Data history was actually so brilliant that I thought worthwhile to reproduce it here.
It all started with simple surveys and regression methods. Over a 100-year period, volumes kept increasing and modelling methods have coped with this evolution.
Yet, with research at Microsoft showing that prediction gets better when volume increases, one may wonder whether a need still exists for more complex models. And in general, are models produced by academics still relevant given that academics only have access to small samples (compared to those of the industry) ?
I don’t have an answer to this question. I can only stress that the gap is growing between academics and practitioners.