Basically, it is blindly tasting three samples, and trying to tell which one of the three is different. (The other two are the same.) So, if you are testing a new supplier, you can give the subjects two samples from the old supplier and one from the new supplier, and see if they can pick out the odd one with any kind of statistical significance.

The key factor is that it is not attempting to measure "which tastes better", but rather, "which tastes noticeably different".

It may use ingredients that score better on certain metrics (low sodium, etc.) but these don't mean"better quality" in the sense that these consumers are interested in (the taste that they've come to like.)