The Victory Lab (44 page)

For too long, Bird thought, field organizers had collected data on the volume of contacts their teams had made largely for the purpose of
impressing campaign higher-ups in memos or bragging to the media about the fearsomeness of their operations. But the metrics-obsessed Obama campaign realized that these figures were not particularly insightful. Who cared how many calls you placed if most of them went unanswered? Or how many doors you knocked if they belonged to voters outside your target universe? When forced to organize in barbershops and salons, however, Bird’s field staffers were able to efficiently deploy volunteers only because they adjusted their voter contact data to account for the fact that canvassers were reaching people not at their homes but at other places where they gathered. “There’s a myth that if you’re using really smart data for targeting, you’re therefore marketing and you’re not doing people-focused, relationship-based organizing,” says Bird. “It hopefully helps our organizers and volunteers talk to the right people, and gives us a better sense of who we’re talking to when we call them.”

When a field organizer like Bird touched down in a state, one of his first jobs was to divide his area so it could be manageably canvassed. Typically campaigns did this along familiar lines, the indelible boundaries of counties and legislative districts for which both reliable statistics and established lines of political responsibility already existed. But the Obama targeting desk decided that field staff no longer needed to respect the cartographic primacy of parties and political institutions. Wagner believed that, through individual-level modeling, he had successfully disaggregated Ohio into a roster of people who could be reshuffled around the campaign’s needs, like its national objective to have such omnipresent field offices that frequent visits would become a matter of routine for volunteers. (Obama would eventually open nearly one hundred of them in Ohio, including five in Cuyahoga County alone.) Wagner’s analysis allowed Bird, who would say things like “You have to let the ground tell you when your targeting is working and when it isn’t,” the ability to survey his turf with an eye for the clumps of the people Obama needed to reach rather than the terrain itself.

Bird settled into the main office in Columbus with Wagner’s vote goals spreadsheet and imagined the way he would want his Ohio to look.
He had helped to develop the Obama campaign’s field organizing system, a pyramid structure with volunteer neighborhood teams at the bottom. Already the staff organizers responsible for building and managing teams in their regions would have a new tool supposed to transform volunteer recruitment. The targeting team had developed a model that would predict which voters would be most likely to volunteer: those with high support and turnout scores who contacted the campaign were aggressively pressed to contribute their time. Instead of seeing volunteerism as an act distinct from voting, Obama’s data team tried to quantify it on a continuum of political behavior. Once a volunteer was on board, Bird would monitor his or her progress along a series of actions that he thought of as adding up to a “leaders test.” Those who earned their way into the volunteer hierarchy joined a neighborhood team, each with a lead person and at least three deputies given specific responsibilities: a phone bank captain, canvass captain, a data coordinator responsible for making sure field contacts promptly reached the VAN. (In some places, Bird added a faith liaison to local church communities.)

The goal was to have these volunteers “own their turf,” as Bird explained it to them, and that required the neighborhoods to be drawn the way people thought of them. Relying on a targeting-desk analysis of population density, Bird classified every precinct in the state as either urban, suburban, exurban, or rural, and made sure that neighborhoods did not mix categories so that team leaders could master their area—the best way to canvass the housing projects on one side of Youngstown’s freeway might be entirely different from the the way to canvass tract houses on the other.

But even though Bird asked volunteers to help generate “best practices” for canvassing neighborhoods, he would not be relying only on their anecdotes and guesswork about efficient voter contact. In early summer, Wagner had tapped the Matrix to develop yet another set of individual-level modeling scores predicting how easy it would be for volunteers to reach a given voter. One, called a canvassability score, measured the odds that a voter would answer his or her door when knocked; a callability score
pegged the likelihood he or she would answer the phone. In Indiana, Matt Lackey would go even further, predicting what time of day a given voter was most likely to respond to a call. If you have volunteers in a field office on Tuesday at noon, who do you put on their call sheet to increase the odds that they will get through to people then?

AS OBAMA’S STATE DIRECTORS
settled in to their new offices for the general election, several of them were shocked to find a scary number lurking deep in their summer polls. Voters ages eighteen to thirty-five were splitting between Obama and McCain. These young voters were supposedly the bedrock of the Democratic nominee’s coalition, having provided the margin for some of his most important primary-season wins. Splitting them with McCain would effectively guarantee Obama’s failure, and worry from state headquarters quickly found a home in Chicago, too. Academic researchers had begun to suspect that changes in telephone culture had made it nearly impossible to successfully locate and reliably sample young, mobile voters, but the political world—with its appetite for rapid, cheap surveys—had made few adjustments to compensate. Grisolano decided it was time to solve that problem. He hired Anna Bennett, a pollster outside the triumvirate of survey takers responsible for Obama’s polling operation. They had been using the standard method of randomly calling numbers, screening for those likely to vote, and adjusting the sample to ensure it reflected the expected composition of the electorate. Bennett instead specifically went after a sample of eighteen- to thirty-five-year-olds, dialing only mobile phones, a far more costly method. Grisolano welcomed her findings with mixed feelings: Obama’s strategic position was safe—he had an overwhelming lead over McCain among younger voters, after all—but the results also showed that the traditional polling method was useless for this demographic. Only 27 percent of Bennett’s respondents possessed a landline.

The campaign leadership sent the Bennett poll out to state directors, and the reassuring top-line number had the intended palliative effect. But Bennett’s research highlighted a broader challenge Obama’s campaign faced: one of its core constituencies could not be easily reached through traditional campaign methods. The voter-file managers went out and bought mobile numbers from commercial vendors, to help ensure that polls and ID calls came out of a realistic sample of the population. Once those phone numbers went into the database, they became grist for the constantly churning algorithms, which found that just having an available mobile number had a predictive power. Cellphone ownership was a proxy for other demographic groups that were established parts of the Obama coalition—the young, the mobile, minorities—and that made identifying cellphone numbers doubly valuable for members of the data team. They eventually took all the phone numbers they had on file, stripped them of the attached voters’ names, and sent them out to a commercial database with a request to flag those numbers that belonged to cellphones. (In part because privacy laws regulate mobile phones differently from landlines, they are coded separately.) The data desk now had another variable to help it predict who would be an Obama supporter even if—as was often a problem with caller-ID-equipped cellphones—the person on the other end never answered a call from a volunteer phone bank or professional call center to confirm that informed hunch. The episode was typical of the empirically minded Obama campaign, which was often so ambitious in its troubleshooting of practical concerns that staffers ended up solving problems they didn’t know they had.

But the conundrum of how to reach young voters never went away. After working in four states during the primaries, Mike Moffo had invented his own job in Chicago, overseeing “special projects” for the field department, a portfolio he described as “all this cool stuff you want to do as a field director if you have time.” Moffo focused on ways to draw new voters into the process, especially young and minority ones, and became known as a font of slightly bizarre ideas that pushed the bounds of where politicians
usually took their messages. He studied guerrilla-marketing techniques used to sell albums and liquor, and pored over Big Ten football schedules to find opportunities when two battleground-state schools were playing and it made sense to charter a plane to fly over the football stadium with a banner listing a phone number to call for early-vote information. When modeling scores showed that the presence of a teenager in the household was a variable that made a swing voter more likely to support Obama, Moffo and Michael Organ, a member of Grisolano’s media team, schemed to buy space within video games. Soon the early-vote number appeared as advertising in hockey and NASCAR video games; higher-ups said violent games were off-limits for Obama’s image.

Each time Moffo had a new idea, he had to shop it around the different departments and find someone to sponsor it. “If you have the purse strings inside a campaign, you’re somebody,” he says. While waiting to be somebody himself, Moffo tried to build credibility through self-branding. Moffo was obsessed with giving his projects code names, dubbing a canvassing-list cleanup tool “Houdini” and a pre-election volunteer rush “November Rain.” He called his new one-person unit “the Jigsaw” and designed a logo that he put on all the presentations he developed for others in the campaign.

As he looked for patrons for each project, Moffo worried that the campaign’s willingness to innovate—and to push the stylistic bounds of politics, especially in the service of defining Obama as the cool candidate—was doomed to clash with Chicago’s data-centric culture. This conflict was embodied at the top, in Plouffe, who despite his frugality and insistence on proven outcomes often found his ledger flush with
cash he considered “Monopoly money”—ready to be freely dispensed to subsidize a culture of inquiry.

In December 2007, Obama’s online team decided to test one of the staple elements of its Web splash page: the button accompanying a small form where visitors were asked to submit an e-mail address and ZIP code to join Obama’s mailing list. The site had been using the words “Sign Up” on its buttons, but for the experiment programmers created three
variants—“Learn More,” “Join Us Now,” and “Sign Up Now”—and randomly assigned them so that visitors would each see different designs when they arrived at the site. After more than three hundred thousand visits, analysts were able to see that “Learn More” yielded by far the best results, nearly 20 percent more sign-ups than the standard “Sign Up” button. (“Sign Up Now” did slightly worse than the original.)

From that point onward, nearly every time the campaign reimagined its Web splash page or asked for money via e-mail, designers created several options so they could run a randomized experiment comparing them. These A/B tests, as they are known, were straightforward: whichever treatment received more clicks or brought in more money worked best, and would go out to the wider population and contribute to a new set of best practices about winning attention online. There were even glimmers of insight into human psychology; in Iowa, website buttons that said “confirm your caucus location” drew far more clicks than those that said “find your caucus location.” (“If you say ‘find,’ people say, ‘I’ve been caucusing my whole life,’ ” says Organ.) “Everyone was just ‘metrics, metrics, metrics, metrics,’ ” says Moffo.

He said that with a little less awe than others in headquarters. Moffo was part of a cadre of Obama staffers who saw metrics having a distorting effect on the campaign’s priorities; because digital communication was so easy to randomize and measure on a minute-by-minute basis, those who bought online advertising would always have the upper hand in budget debates. It was simply easier to demonstrate results there than in mail or broadcast ads, and certainly easier than in the nontraditional political media that caught Moffo’s attention. “The online department doesn’t understand this stuff because they understand opening rates on e-mail and fund-raising rates,” says Scott Goodstein, who had bounced between music promotion and political campaigns before mixing the two as cofounder of the advocacy group Punk Voter and its related fund-raising series Rock Against Bush. “That wasn’t the world I came out of. I understood how you get people to go to a concert or buy a record because they’ve heard of the band.”

In 2006, Goodstein had been one of four organizers behind a massive antiwar march on the National Mall, the crowd rallied almost entirely with digital methods. The next year he joined Obama’s new-media team as its external online director, which made him responsible for Obama’s digital presence just about everywhere other than his own website. Goodstein established Obama’s presence on social media sites like MySpace and Twitter and produced original content for them. By the general election, Goodstein felt there wasn’t much more work for him to do on those sites. “You didn’t have to convince people to watch a Barack Obama video on YouTube. My initial job was done,” says Goodstein. He began to wonder how he could translate his instinct for word-of-mouth communication to reach those young voters, especially non-college-educated ones, who were not active consumers of online political media. “How do I get the offline people?”