Led by an M.I.T. political scientist named Ithiel de Sola Pool, the chairman of Simulmatics’ research board, Greenfield’s scientists compiled a set of “massive data” from election returns and public-opinion surveys going back to 1952, sorting voters into four hundred and eighty types, and issues into fifty-two clusters. Then they built what they sometimes called a voting-behavior machine, a computer simulation of the 1960 election, in which they could test scenarios on an endlessly customizable virtual population: you could ask it a question about any move a candidate might make and it would tell you how voters would respond, down to the tiniest segment of the electorate.