Unions are “crunching numbers,” according to this Post report.
The AFL-CIO wants to spend several million dollars on a mailing to its members. Traditionally, a focus group of about 12 people would be convened, and a series of different messages would be tested on them. The message the group liked best would be used for the broad-scale mailing.
Under the model Podhorzer employs, random samples of 20,000 (or so) union households each receive a different piece of mail containing messages about the two candidates. One sample group receives no mail at all. Then, the AFL-CIO conducts a poll of roughly 1,000 people in each subgroup to test how the mailing affected their vote, rather than whether people liked or disliked the message.