Stanley Greenberg in The American Prospect:
The campaign relied far too heavily on something that campaign technicians call “data analytics.” This refers to the use of models built from a database of the country’s 200 million voters, including turnout history and demographic and consumer information, updated daily by an automated poll asking for vote preference to project the election result. But when campaign developments overtake the model’s assumptions, you get surprised by the voters—and this happened repeatedly.
Campaign manager Robby Mook and his team believed that identity politics, demographic trends, and Trump’s temperament would be enough to win, so they could avoid confronting the “trust problem.”
Astonishingly, the 2016 Clinton campaign conducted no state polls in the final three weeks of the general election and relied primarily on data analytics to project turnout and the state vote. They paid little attention to qualitative focus groups or feedback from the field, and their brief daily poll didn’t measure which candidate was defining the election or getting people engaged.
The fatal conclusion the Clinton team made after the Michigan primary debacle was that she could not win white working-class voters, and that the “rising electorate” would make up the difference. She finished her campaign with rallies in inner cities and university towns. Macomb got the message. “When you leave the two-thirds of Americans without college degrees out of your vision of the good life, they notice,” Joan Williams writes sharply in White Working Class: Overcoming Class Cluelessness in America.