How Obama’s data scientists built a volunteer army on Facebook

No matter how good your social media team is, the chances are it’s never done anything like this. Rather than just using Facebook (s fb) as a channel for posting messages and tracking its followers’ feelings, the Obama for America data science team turned social media into a tool for efficiently recruiting the human resources it needed leading into the election’s home stretch.

The key was a model for determining who among its followers were the best messengers, who they might be able to persuade, and what actions they might be willing to take. So, rather than blast all of President Obama’s 30 million Facebook fans or 20 million Twitter followers with the same plea for cash or neighborhood organizers, the campaign was able to make informed decisions about whom it asked for what, and how it asked them.

Rayid Ghani, Obama for America Data Scientist
Rayid Ghani, Obama for America Data Scientist

During a recent interview, Obama for America Chief Scientist Rayid Ghani compared his team’s social media approach in 2012 to the shift in web content from reposted print material to material designed for the web. For many organizations, he said, the prevailing strategy is “‘I used to use email, and now I’m just going to put the same information on a Facebook page.'” However, the president’s campaign used an abundance of online and offline data in order to hyper-personalize messages and get the most bang for its buck in terms of outreach.

Essentially, Ghani explained, the campaign was able to match up supporters’ friends against voting lists and determine how it should approach supporters to reach their friends. If someone was going to spread a message to 20 people, the campaign wanted to ensure they reached 20 people most likely to take action in some way. Because Ghani’s team had done so much work integrating its myriad data sets into a single view, it was better able to decide who could be most easily persuaded to vote for the first time, to donate money, to get active knocking on doors or perhaps even to switch sides.

That it was coming from friends rather than the campaign was critical to the strategy’s success, too. “The more local the contact is,” Ghani said, “the more likely [people] are to take action.”

Changing times call for changing communications

The effort to build an intelligent system like this was necessary because younger voters’ means of communication had shifted so greatly even since 2008. Then, Ghani explained, the Obama campaign relied primarily on phone calls and neighborhood canvassing in order to reach people. Now, many young people aren’t reachable on landlines at all, but they are always connected to some form of social media.

Not that the president’s campaign abandoned those traditional methods, though. Ghani said many older voters are still best reached via non-digital means, and even when his team was using volunteers to reach those demographics, it suggested they give them a call or talk in-person. It really was all about what was most effective for each individual.

Four years from now, however, Ghani sees a very different picture again in terms of how campaigns will reach their voters. Younger people will be even more difficult to reach via traditional telephones, and they might not even be watching a lot of over-the-air or cable television. On the other hand, he noted, older people will probably be more engaged online and perhaps with mobile apps, as well.

“Reaching [people] through the channels they’re most engaged in is going to have to become mainstream [in political campaigns],” Ghani said.

However, he reiterated, because there’s only so much money to spend on any given medium, it’s more important to inform those interactions with voter intelligence than just to make them. Speaking about increasingly targeted online advertising, Ghani explained, “You don’t want to waste impressions on people who are not in your target audience. … If you’re lactose-intolerant, there’s no point in showing you ads for yogurt and dairy products.”

And even on the campaign trail, data was making a difference. Ghani’s team wasn’t writing speeches or managing communications, but it was helping the people who did those things do them better. For example, he explained, Matthew Rattigan, an analyst on the team, built a tool for looking at the coverage of speeches in local newspapers so it could break down by geographic region how people reacted and which parts were quoted most. Speechwriters were therefore able to see how the messages they wanted to convey were actually the ones that were covered.

Very hard work, very short window

It’s pretty amazing to think about how much the Obama for America Tech team accomplished — not just Ghani’s data science team, but also CTO Harper Reed’s unit and the team led by Chris Wegrzyn responsible for building the analytics and data infrastructure — when you consider the short time frame in which they had to do it. Ghani considered himself lucky that he and his team inherited a handful of staff and techniques left over from the 2010 mid-term election, so the 2012 team didn’t have to begin with the cupboards bare when it got to work in mid-2011.

But, still, it had just a year and a half to recruit a team of data scientists (“finding people who were qualified was extremely hard,” Ghani said) and to get some foundational data-management systems in place. The campaign’s targeted outreach efforts on Facebook, for example, only got up and running in August following an extensive effort to integrate the campaign’s myriad online and offline data sources.

“Our biggest challenge this time around was getting all the [online and offline] data together in one place,” Ghani said. “We were basically a political campaign with the problems of a large enterprise.” However, he added, corporate data architecture is an ongoing concern — they don’t build systems for one-off jobs then abandon them.

Given all that work, he thinks it “would be a shame to let it go” when the next presidential campaign team is built. Not that what it did will still be totally relevant four — or even two — years from now. Data sources will change, as will technology, but having something in place is better than nothing.

This is especially true considering how little budget campaigns typically have between elections to focus on technology. Absent an abundance of money and the right people during election season, Ghani said, “You need to make sure you have people whose job it is to be thinking about this stuff all the time.

That’s because although the work Ghani’s team wasn’t enough to win an election on its own, it was very important. And in tight elections, any competitive edge is worth having (while failures can have harsh consequences). Reaching the right voters at the right time with the right message will become even more important in future elections, he said, and “we only scratched the surface of it.”