I wonder if the Cheesecake Factory and Applebees use A/B testing with their menus the same way Netflix uses A/B testing? There is probably a huge demand for data driven decision making for restaurants rather than these spares studies mentioned in the article which most likely are anecdotal. Simply have four different menus with the same items and look at the analytics provided by point of sale system. That is a fun start - up idea.

Most of your favorite restaurants in places like San Francisco, Portland, and New York City engineer their menu by removing least popular menu items once a month or once a season and replace them with what they think their customers currently demand at a price to insure that the restaurant is running at the point of highest efficiency. There are other considerations such a cost of product but most restaurants increase overall demand by using analytics from the point of sale system and the most profitable dishes are not always the most popular. Engineering a kitchen -- and dining room for that matter -- to be efficient without cooks tripping over each other is another whole discussion nonetheless even this ties into what can be added to the menu.

The two biggest mistakes an independent restaurant can make is hard coding a menu and putting the word in Bistro or Basserie in the restaurant name -- the restaurant can't evolve after that. Have any of you been to La Folie on Polk and Green in Russian Hill? That restaurant first opened as a casual French Bistro but the demand was for their caviar by the ounce and Tournedos Rossini which oddly enough was not only their most popular menu item but also not on the menu. So in the mid nineties the average customer was spending $150 - $300 instead of $40 per person.