In the movie Moneyball, Brad Pitt builds a winning baseball team using predictive analytics and takes the Oakland Athletics all the way to the playoffs. Obama for America has followed the same strategy leading up to its own championship game -- today's election.
The presidential election is shaping up to be a litmus test for Obama for America's all-in commitment to predictive analytics. Team Romney is using data mining for individualized ad targeting. They have micro-targeting experts on staff, but not the team of data scientists recruited by Obama for America. "It's not quite the same skill set," says Patrick Ruffini, president of Engage DC, a political consultancy.
Meanwhile, Obama for America has applied predictive modeling broadly. "They are taking it up a notch in data mining all aspects of the campaign, from online behavior to modeling door knocks to taking down all of the information," Ruffini says.
Analytics by itself can't win an election, but the careful application of predictive modeling at all levels of campaign strategy could make all the difference in a close race like this one, experts say.
If Obama wins, expect both parties to double down on data mining in the next election, with more extensive use of real-time databases that continuously receive updates from the field and the use of analytic models that adjust campaign strategy accordingly.
Regardless of the outcome, however, campaigns are likely to pursue even greater integration of online and social media data and will expand the use of mobile technologies. You'll see more canvassers knocking at your door with devices like Square to take credit card donations on the spot, along with mobile apps that tell the volunteer how much to ask for based on previous donation history. "Square lowers cost, increases productivity and converts supporters into donors more quickly," says Patrick Hynes, president of Hynes Communications, an advisor to the Romney campaign.
"Facebook and to some extent Twitter have become a repository for consumer data that's probably unequaled in history in the sense of consumer intent and preferences and political affiliations," says Ruffini. All of that is now feeding into the hopper to drive both tactics and strategy.
Analytics permeates every aspect of the Obama campaign, right down to the auto-dialer get-out-the-vote programs that decide who to call and give volunteers a script with a message designed to have the highest probability of moving each micro-targeted segment of voters to action. One Obama volunteer said that a program that clearly identified itself as part of the Obama team's predictive analytics effort made the calls for her last night. (At this late date, however, it may be having little effect: She reported that most people aren't picking up their phones.)
"As the elections get closer, predictions from the modeling become as accurate as the field survey polls" -- or more so, says Cathy Duvall, director of public advocacy and political director at The Sierra Club, which has made a substantial commitment to predictive analytics in its own messaging and get out the vote drives this year. Analytics, she adds, don't rely on people telling the truth in surveys.
That's exactly what's happening in the Obama campaign right now. In an ABC news report last night, Diane Sawyer mentioned that the Obama campaign isn't using some of its traditional polling partners in the same way campaigns have in the past.
Obama for America's approach is experimental, however. This election will be a referendum of sorts, not just on the results politicians can expect to achieve but on just how broadly analytic models should be applied. Obama for America is running the world's first top-to-bottom, data-driven campaign. If Obama wins, it will forever change the rules of the campaign game.