It works well for me. I'm good at it. I find it very easy to get a "lay of the land" in terms of groups, structures, power dynamics. Pretty easy to infer group preferences from there. However, your question here is completely besides the point. Everything we've been talking about so far has been about general group
social discomfort. When I want to find out individual preferences, I'll focus on the individual. In social situations, inferring individual preferences from the whole makes no sense -- it's inaccurate, inefficient, and generally a waste of energy. My social "lay of the land", when it isn't instinct, is a collection of individual data. It doesn't work the other way around.