How would you even assess the profitability of a small recipe change? You'd only notice by the time a customer comes back and there are so many variables (and time) going into that I can't imagine you could really get some signal out of your a/b-test of a few drops of chili oil in the soup.
If you have 10 restaurants you could do A/B tests.
This is massively prevalent for food science and prepackaged goods (through panels and such; usually not to the end customer). Seems like restaurants are catching up.
I've also seen people test two ingredients, one cheap and one not. If their sample can't tell the difference, it flies. The problem is, we're each sensitive to different things. So while I may not notice the difference between a pricey and mid-range whisky, I will notice when you sub out the Chartreuse. Meanwhile, my neighbor will likely be the opposite. Rinse and repeat this process and you get the beige this process predictably produces.
