The Obama Campaign's Legacy: Listen, Experiment, and Analyze Everything
BY Nick Judd and Micah L. Sifry | Wednesday, November 21 2012
The big story of the 2012 Obama campaign was not just that staffers were able to weave together information about voters using data integration in ways that had bedeviled the campaign in 2008. Nor was it the campaign's ability to test nearly every tactic, from email subject lines and landing pages to the scripts that volunteers read from as they went door to door. If any one engine powered the campaign down its road to victory, it was the system that turned every voter, field staffer and grassroots volunteer into a political signal — and the mix of technology and analysis that allowed Obama's Chicago headquarters to find those signals among all the noise.
Eighteen months ago, we predicted in these pages that this election would be all about data, stupid, and that President Barack Obama's campaign might use Facebook data as a force multiplier. Now that campaign headquarters has been emptied and many of the staffers who worked on the nuts of bolts of the Obama machine are starting to talk, they tell techPresident that very little about the campaign was new except for its sheer size. What made it different, they say, was the discipline of its staff and the efficiency of machinery improved and retooled since 2008.
"This campaign had enough data to be able to make differentiated strategy decisions at small geographic levels," one campaign staffer told techPresident. "The unsung heroes were all the team leaders and volunteers who were entering data nightly, and then strategizing with the data coming back from HQ."
From the campaign's data-driven approach to ad buying to its emails, the driving principle, based on interviews with several people who worked in or around the campaign, was to take an evidence-based approach to person-to-person contact. The core of the campaign was not flashy or even particularly innovative except in the willingness of senior staff to listen to numbers people rather than to consultants acting on old-fashioned political intuition.
"The power emerged this time from the integration of a ton of different things," said the same staffer. "It's not a sexy story. Not super innovative, but being highly disciplined."
From a certain point of view, the only truly new thing in the 2012 Obama campaign was the ability to get all of the campaign's gears to work together more smoothly. The door-to-door ground game was a legacy of 2008. Obama for America also stood out in 2008 for a structure that placed senior digital staffers on the same level as other top staff rather subordinate to another shop within the campaign, something that many campaigns emulated in 2010 and again in 2012. In this election, OfA merely put more resources into an area — technological campaign infrastructure — that the evidence indicated would be useful to help them win. As chief innovation and integration officer, Michael Slaby sat at the campaign's highest level, with a chief technology officer, chief information officer, and chief analytics officer all reporting to him.
Narwhal, the system that allowed the campaign to join together data from multiple databases at once, is a prime example of an improvement heralded as an innovation. In 2008, the campaign also sought to connect the data it held in its voter database with the information supporters were giving them online. But integrating that data in real time was beyond the skill and tools of the 2008 campaign, so they relied on overnight batch jobs to do some of the things that Narwhal did in 2012 with more speed and efficiency. Dashboard, the social network for field organizing (that we took a close look at here last summer), glued together a number of tools — such as social groups, events, access to the VoteBuilder voter database, and click-to-call functionality — that the campaign already had in one form or another going into 2012. And the art of the perfect fundraising email has been a topic of discussion in Democratic campaign circles since the days of Howard Dean.
"There were analytics types all over the campaign," said Amelia Showalter, the campaign's director of digital analytics. Besides digital, there were separate analytics teams dedicated to modeling, the battleground states, polling, paid media, finance and communications.
Some of these teams did work that was done because the campaign had the capacity to do it, not necessarily because they knew it would have value. The analytics team focused on communications not only analyzed earned media, it sought to give the campaign's principals insights into real-time coverage and commentary that would in theory allow them to get ahead of the news curve.
"After a big speech or after the debates, you can run an analysis across the entire Internet to see what the important lines are coming out of the debate, what was covered, what wasn't," a top staffer told techPresident. "So you can kind of pre-empt what was coming out of the cycle."
Because the campaign's digital team — separate and apart from the work of CTO Harper Reed, whose team was focused on technology infrastructure — was so large, it needed its own team of analysts for emails, online ads, social media and online fundraising.
"There was constant testing," Showalter said. "A lot of people were responsible for that culture ... a culture of experimentation. If we had an idea that was kooky, we would test it to see if it would work."
For example, for a fundraising email, the digital department decided to try emphasizing text with highlighting that was ugly on purpose. It outperformed other emails, so the campaign kept using it — the formatting trick seemed to draw the eye. When the novelty wore off and that tactic stopped performing better than other ones, the campaign dropped it and moved on.
That culture of experimentation extended to the amount of responsibility passed on to volunteers. With offices in San Francisco and a later expansion to New York, a campaign team called Tech for Obama created room for volunteer designers and developers to write code for the president. Volunteers built tools like Trip Planner, which matched people driving into swing states like Nevada from "safe" states like California with other volunteers who needed a ride, and several of the graphics that the campaign distributed on social networks. Those volunteers, as well as the engineers that worked in Chicago, are now intimately familiar with the needs of political campaigns and the experience of working on an election, which may create a new class of disruptive actors in politics the way the Obama 2008 campaign did by training volunteers in grassroots organizing.
The campaign's ability to scale increased its ability to facilitate one-to-one contact, which increased the amount of data coming back to the campaign, which allowed headquarters to more efficiently direct additional one-to-one contact.
The 2008 campaign made 30 million phone calls to voters, according to Daniel Kreiss' book "Taking Our Country Back: The Crafting of Networked Politics from Howard Dean to Barack Obama." In a campaign memo sent the weekend before election day, campaign officials claimed they had already made 125 million calls to voters or door-knocks in 2012. In an interview with Politico on Tuesday, campaign manager Jim Messina said the campaign registered 1.8 million people "on the doors" and another 1.1 million people online. He also said that the use of targeted sharing, the program that combined Facebook data on users with internal data on voters among those users' friends to suggest which content those users should share, allowed the campaign to reach more than five million people "directly through their Facebook world and people that they knew."
Some observers like academics Zeynep Tufekci and Dave Parry have raised concerns that this amounts to social engineering on a massive scale, rendered all the more insidious because campaigns have tools of persuasion that Edward Bernays could only dream of: a long list of people who know members of target audience personally and are willing to lobby that audience in person.
The Obama campaign's targeting operation may seem to hint at an Orwellian political future where individual citizens are helpless before a class of elite manipulators, pandering to their every stated preference. The current reality is a little less dire. The Obama campaign didn't get much use from commercial data like magazine subscriptions or purchase histories — the only really useful outside data was largely demographic.
"I mean, the best data for us was things that we collected at the doors," Messina told Politico. "You know, there was some commercial stuff that every campaign, including the Romney campaign, purchases. But you know, the truth is, the more we learned about data, the more we learned how important the connection was. The door-knocking, having a real conversation with people on the doors, that really mattered."
The campaign also relied heavily on information that is public record, like voting histories. None of this is functionally different from the kinds of insights that campaigns have used for years to target direct mail, and people who worry that politicians are learning how to tailor their message to their audience based on internal research are decades late to the game.
This practice is merely becoming more sophisticated, and although one could argue that political media must now step up its own game to keep pace in order to expose when targeting turns into pandering, that's already starting to happen. Targeted ad runs that would be hard to uncover because they only reached a few thousand voters, digital strategists tell techPresident, are also likely to cost more per potential vote than a campaign should really spend and not reach enough voters to be worth the effort. The larger the group a targeted campaign reaches, the more likely someone in that group would be to participate in something like ProPublica's Message Machine project, which collected and sorted various targeted emails from throughout the campaign. Forwarding an email or taking a screenshot of an ad is a lot easier than scanning a piece of direct mail or forwarding a physical palm card to a local newspaper.
But the Obama campaign was able to pick the most effective persuasion tools in its arsenal to an unprecedented degree. We now know, for example, based on our conversations with Obama staff, that decisions about where to send each of the four key principals — Obama, Vice President Biden, Michelle Obama and Dr. Jill Biden — were tested against the campaign's data modeling in order to make best use of their different abilities to connect with voters.
Being too obvious with targeting, or missing a target, so to speak, can hurt rather than help. In September, Yale political science professor Eitan Hersh released the results of a study in which people were asked a series of survey questions, presented with a narrowly targeted piece of direct mail from a fictional candidate, and then asked to respond to that message. Each piece of mail was modulated to speak in no uncertain terms to one of a few groups, such as Latinos or evangelical Christians — in other words, to pander. When someone in one group saw a piece of mail clearly aimed at someone in another, they generally indicated they were less likely to vote for the candidate sending it. But the study also found that when someone indicated they believe their fortunes rise and fall with a group, their opinion of a candidate improves when they see a message pandering to that group — even if they're not necessarily part of the group.
"Of all the results in the paper," Hersh told techPresident earlier this month, "what I think we're really sure about is people don't like when politicians target an identity that they don't really hold."
Hersh cautions that his study asked people for their reaction to a hypothetical situation, rather than observing how people reacted to an actual candidate. He was also basing the study on pieces of direct mail he had obtained that made obvious appeals to narrow groups, rather than persuasion designed to be more insidious. The practical hazards for a campaign may be lesser and, as a result, the ethical hazards greater.
None of the Obama staffers we spoke to had a clear answer on whether what they had done for the campaign was social engineering of a troubling kind. What they were clear on, including people for whom the campaign was a first introduction to politics, is that it has changed their own view of elections for the better.
"I felt better about politics," said Will St. Clair, a programmer who worked on the software that allowed OfA to mine a website visitor's Facebook friends and come back to the visitor with content they wanted those friends to see. "For it to go from something that was an abstraction to something that you can work on and give it your best in conjunction with other people on it, really make a difference in all sorts of ways -- it's been very good."