I've worked on this in the past too (not as rigorous as it sounds you are). My main problem: how do we define our optimum function, and how do we quantify the variables going into it? How do you do it? I assume you're not sitting there and tasting, comparing and scoring 25 salad samples every 3 days...