I have been attempting in oil and gas to educate and describe advanced data analytics for the past five years. The understanding is usually based on the individual's experiential learning and academic education. For example, Dr. Lev Tabarovsky, Technology Fellow at Baker Hughes further enhanced my thinking.
During a recent conversation with a colleague who was my manager at Baker Hughes as well, Dr. Mario Ruscev, currently the Global President of Products at Weatherford, he articulated probably one of the best descriptions I like to share:
"It has become very fashionable to also discuss Analytics use in the E&P world lately. Analytics are really a set of algorithms that empirically define and/or find relations, distances, similarities between multiple variables and/or dimensions events and allow you then to obtain responses when physically derived models cannot be done or struggle. Plugging data into Predix or Watson and expecting an answer does not really work. First you need to define what question you want to answer, then you know your events and get together a set of mathematicians and subject matter experts to find the type of algorithms that will work and optimize your solution if one is found. Instead of using a closed eco system like Predix or Watson we rather form a small team of Data Scientists (in the old days we called them theoretical scientists) to develop the solutions."