You do not understand polling and market research.
People give their opinions on a show, usually a very fav, somewhat fav, neutral [volunteered, rarely read out loud as an option], somewhat unfav, very unfav. They are likely then asked to rate aspects of the show on a scale of zero to five or one to ten. The individual ratings are opinions. Once placed in these categories, they become quantitative data. An open-ended question such as "what did you like most about this episode," is qualitative data and those opinions would be classified into several categories to make them quantitative.
Now, I'd bet the networks do dial tests of their shows. They give each participant a device to hold with a dial on it. You turn it to the right when you like stuff and to the left when you do not. This technique is used in political campaigns. I'm sure you can see what were then, live, dial tests done during the presidential debates. But a lot of commercials and television programs are tested this way. This is how consultants learned that people are less patient listening to female political candidates speak because people turn the dial to the negative sooner with female candidates than with male candidates. That was one of the famous findings of this research technique.
But, if you are a Nielson family, you write in detail who is watching what program in your household. The toughest demographic to get in front of a television or on the phone is young men. This is why so many commercials are geared to them via music or visuals.
The networks DO know which demographics watch which programs. The most famous case of how this data is used was back in the day with the show "Lou Grant." Lou Grant had good, solid ratings, but CBS cancelled it because the demographic that watched it was "too old." Big Bang Theory still solidly wins the demographic they are looking for, which is folks under 50 years of age.
Advertisers drive a lot of this research.