In the process of product optimization, food engineers alter a litany of variables with the sole intent of finding the most perfect version (or versions) of a product. Ordinary consumers are paid to spend hours sitting in rooms where they touch, feel, sip, smell, swirl and taste whatever product is in question. Their opinions are dumped into a computer, and the data are sifted and sorted through a statistical method called conjoint analysis, which determines what features will be most attractive to consumers. Moskowitz likes to imagine that his computer is divided into silos, in which each of the attributes is stacked. But it’s not simply a matter of comparing Color 23 with Color 24. In the most complicated projects, Color 23 must be combined with Syrup 11 and Packaging 6, and on and on, in seemingly infinite combinations. Even for jobs in which the only concern is taste and the variables are limited to the ingredients, endless charts and graphs will come spewing out of Moskowitz’s computer. “The mathematical model maps out the ingredients to the sensory perceptions these ingredients create,” he told me, “so I can just dial a new product. This is the engineering approach.”