This is one of the reasons why I appreciate Netflix. Instead of a bunch of shows that everyone considers average, it's taken some risks and created shows that some people consider great and others dislike. I'll take that over a bunch of lukewarm shows anytime.
This is a great example of how Netflix's model works well for "long tail" content. They can produce and keep shows around that have a different or smaller audience than traditional networks can.
They started with Game of Cards and Orange is the New Black which are the strongest examples in history of creating content based on what data tells you people want, because those were their first few series and they couldn't afford to have them flop.
Nowadays, they make such a wide range of shows, they can afford to sometimes just take a risk on giving talented people money and hoping they make a great show that finds an audience.
One series I've enjoyed recently is The OA which is definitely not a box-ticking exercise of a series - the complete opposite. I'm happy it's been renewed for a second season, but I can't imagine it's driving any revenue directly.
It's like a homeless man being in a cardboard box and duct tape suit screaming about his mech suit. You just feel bad for him.
It felt like a cheap attempt at being mysterious or deep. If you look at the people behind the show, it's seems like this is part for the course with their work.
However, a data driven mindset should be applicable on deeper levels also, but that probably takes longer to become possible to analyse, and crucially, I belive that it is likely to need a baseline of stylistically optimized content before the value of more abstract concepts can be discerned from the data.
An unfortunate effect of data driven is also that unless you also chase some crazy (possible) trends in the data, you will regress to content with maximal earnings in the short to medium term.
Data doesn't carry visions, at least not by themselves!
That is very much the feeling I got from Stranger Things. It is solid in almost every way, but it also felt like it was written to maximize variables and checkboxes from a user survey. Add characters, references and stylistic choices to get the most 80's nostalgic feeling in the viewer as possible.