Blue Yonder, a SaaS predictive analytics platform, makes some very large claims about the automation of predictive analytics tasks, and refers (to the point of tedium) to the use of similar technology at CERN. There is a world of difference between scientific uses of machine learning techniques and commercial use. The laws of physics are constant (if they are not we really do have problems), whereas customer buying habits and other business variables change constantly. This does not invalidate the excellent technology on offer, but it should introduce a note of caution. Fully automated analytics, where humans do not get an opportunity to perform a sanity check, is full of risk.
Predictive analytics algorithms are not immune from producing nonsense, no matter how sophisticated they are (in fact there tends to be a link between nonsense and sophistication – ask analysts working on predictive models just prior to the 2008 banking crisis). Bayesian methods are simultaneously some of the simplest and most powerful tools available, but they too can generate erroneous models. Caution is the keyword here, and most of us these days are not so easily wowed by scientific smoke and mirrors.
The main features promoted by Blue Yonder include the ability to fully automate forecasting and predictive model creation at specified intervals, automatic data preparation (really!), and the use of probabilistic models for classification and regression (where numbers are predicted). The application of the technology goes well beyond the usual customer focus to machine failure prediction and other less well known applications.
Other claims include the ability for business users to create their own forecasts and predictive models, avoiding the use of data scientists and analysts. While this may appeal to business users, it is in my opinion a little irresponsible. It seems Blue Yonder is joining the growing club of suppliers who somewhat cynically make these claims to make a sale.
All in all this is interesting and novel technology, but the marketing department seems to have lost touch with reality. Certainly worth investigating, but do not leave the future of your business in the hands of an algorithm.