Big Data and Democracy: Lessons Learned on the Obama Campaign
Chief Data Scientist, Obama for America
(Click here to learn about our DataLove event in New York on October 8.)
Obama’s re-election campaign gave me first-hand experience in the potential of data to solve large-scale problems in the political arena, Of course we had a very real deadline and a very specific goal: get Obama elected on November 6. So when it came to analytics, we didn’t have the luxury of examining the data from every angle and find “insights” that we couldn’t eventually act on; we had to work within a very short window. What we learned—aside from how to survive without sleep for months on end—should benefit many organizations that want to use data to increase organizational decision-making, as well as consumer engagement and participation.
“Data mining,” the buzzword, is too limited for today’s data challenges. True, we mine data for the “gold,”—business and consumer insights—but it doesn’t stop there: insights are valuable only when you act on them, or “operationalize” them. We found it was vital to start with a simple data set (such as a user’s past political activities) and a clear objective (such as improving email click-through), and then deploy, test, refine, and repeat. We used machine-learning approaches (algorithms that detect data patterns to learn and improve their own functioning) to constantly evolve and improve our analytics.
We uncovered some interesting “gold.” For instance, we learned there was more than one kind of undecided voter: those who were apathetic, and would probably never vote; and those who were truly undecided, who could be persuaded. By identifying this important second group, our subsequent efforts were more targeted, and more successful.
Also, we found it was important not to start with the data (looking for patterns), but to start with carefully thought-out “business” questions. We embedded analytics into every function in the campaign, from fundraising, to recruiting, to mobilizing volunteers and persuading voters. We worked with all channels—messaging, polling, social media, and TV and online advertisement—and helped every campaign department use data more effectively across these touch-points to personalize all communications and increase their relevance.
David Selinger and I see data similarly: if you use it effectively, you can treat people as unique individuals who change and grow—and thus transform any organization. To do that, you need analytics that evolve as people evolve.
