what is behind "process" it and "further selection" and "refinement" and how big are these "datasets"? These companies ship the encrypted session to you not because they want to.

I agree that they have pipelines for what you are describing but how effective they are at scale and at focusing is the question.