[patrickmclaren]

In the past, when considering the term 'Data Science', I've assumed that this is roughly equivalent to the union of Probability Models, Statistics (with a view towards Decision Theory), CS, and Graphic/Web Design (perhaps throw in some knowledge of Heuristics and Biases too.) The number of individuals with sufficient experience in all of these fields should be significantly smaller than those in any given field alone.

I am aware the the demand for Data Scientists is higher than Statisticians at the present, at least in the US. Assuming that there is a non-trivial amount of employed Data Scientists, one is led to the conclusion that there is fairly large portion of under-skilled Data Scientists.

My question is, assuming the above hypothesis, do Data Scientists, on average, under-deliver, or are job requirements lowballed to avoid failure? First thoughts would be to compare to Quants, or Actuaries.

I have a belief that Data Scientist jobs are created due to the following process: Startup founded => Data collection => Predictive Model Exists => Data Scientist => "Visually confirm" hypothesis and send to marketing department. Obviously, the current order of this chain is not correct. A priori, mere sampling does not simultaneously guarantee regularity and high Fisher information.