The Restaurant Lab

We have collected historical menus photographed at leading restaurants, ranked by La Liste. Each menu is dated, so a dish carries the year it was served as well as the kitchen that served it. Coverage is deepest where menus have been photographed most often, and we are still extending it across regions and back through time.

Flavour harmony scores each combination of flavour notes by how much more often those notes appear together in dishes than chance would predict, adjusting for how common each note is on its own. For example, how much more likely are we to see a third note among dishes that already contain the first two? Ingredient synergy then measures how many highly harmonious combinations the ingredients form, with one note from each ingredient. The more strong harmonies they share, the greater their synergy. Because harmony is learned from restaurant menus, the same ingredients can score differently here than in the Pairing Engine or the Bar Lab. These scores are not a definitive answer on whether a combination works, but an observation based on the data we see.

We model the flavour of each ingredient by analysing thousands of sources, including recipe books, tasting notes, restaurant menus, and food & drink reporting. We find patterns in how they describe flavour and map those descriptions against 150 distinct flavour notes, from smoky to floral to citric, building up a detailed picture of how an ingredient actually tastes based on the collective experience of thousands of cooks and writers. The profiles are then refined by our Flavour Society members, who can adjust anything that doesn't ring true to their own palate. Further details about our modelling method can be found at Flavonomics | Methods.