[brownbat]

> they'd have some inkling as to what's currently in demand and what may be underrepresented.

My working assumption is that the lessons learned from the prediction engine they've built are a qualitative leap over just "what's popular, assuming all of the US is composed of identical viewers." Or even a few demographic divisions.

Eh, I could be wrong though. Maybe this is just a gutcheck, I think Netflix is better at data science than other studios, because they have more experience working with programmers and large datasets.