"actually using that to produce a strategy that worked"
Actually, it can work. Maybe not at very large scale yet, but it's mostly a way to target propaganda.
Categorizing people's "feelings" is actually feasible. I've collaborated with university researches who did it (using Twitter also). Once you can build groups, you can target them with ad-hoc propaganda - and that, like advertising, knows very well how to take advantage of people's weak sides.
After all, Facebook was already using "sentiment analysis" to better target and made more efficient advertising. Probably they are more angry someone else could use their data to achieve it and make money without paying them than anything else.
This was difficult to achieve before for the lack of good datasets, and no way to target the "right" groups/people directly. Something was done obliquely, using other data sets like purchase habits, but they weren't so granular, and you still lacked a direct way to target people. Facebook created both the datasets and a way to reach targets - because people were too naive (or stupid or greed) to understand the risks of helping the collections of so many personal data points.