Lies, Damn Lies, and Data Science

Anecdote, but I have to agree with your call-out the author's paragraph about this is detached from the reality I've experienced. I have hired ~45 data roles: ds, de, and analysts. Not once has a PhD been able to outshine any other, less educated, peers. If anything, those that spent excess time in academia enter the roles with a handicap against them, not in favor, for performance. I will admit I've only hired about 3 with PhDs, but each had major soft skill deficiencies. The analogy is honestly quite offensive with hubris and I would encourage the author to not look down on others.It's my opinion you can only make this evaluation on merit of demonstrated skills and academia is an accomplished program. It's supposed to detail skills or assumed expertise, but that's more than ever with the current state of academia : still very assumptive.

Data science is the practice of producing knowledge rather than building/coding. You can make systems that leverage that knowledge in an automated matter, but when you are deciding what to build, it is not a matter of coding. It is a matter for management. In my experience the sort of arguments a PhD can make fall on deaf ears when management lacks the requisite skills. An organization where the PhD in stats, ML, etc. produces the same outcomes as the code camp hero should look at its management structure before doubting the value of experience at depth.

Considering how hard it is to get statistics right in sanitized laboratory settings, I've always expected most statisticians in private companies to be generating more noise than signal. A lot of people (anecdotal I know) complain that these jobs often come down to 'give the deciders confirmation for their choices'.

This blog post seems to articulate my misgivings better than I ever could. She does conflate APIs and libraries tho which is an unfortunate conceptual mistake when calling out software developers for having poor mental models of their tools.