The Romney Campaign’s Data Strategy
They’re outsourcing it.
Photograph by Nicholas Kamm/Getty Images.
Once a week, members of Mitt Romney’s political department gather in the campaign’s Boston headquarters and pretend that they are in Barack Obama’s war room. Such play-acting has been crucial to Romney’s tactics since he secured the Republican nomination this spring. Shortly after preparing their own vote goals—the state- and county-level accounting of registration, persuasion, and turnout targets that offer the basic strategic foundation of every campaign—Romney’s team drew up a set imagining what Obama’s internal projections would be. In their weekly meetings, they review Obama’s travels to identify the areas they think their rivals in Chicago are making a priority.
At the same time, a new data-science team within Romney’s strategy department sifts through reports on Obama’s broadcast buys as assembled by the Campaign Media Advertising Group. Romney staff code each of Obama’s ads according to their distinctive characteristics: content, style, the candidate attributes and characteristics they’re trying to drive—even the gender of the narrator and at what point the legally mandated “I approve this message” tag is inserted—along with the perceived demographic audience and the markets in which it appeared.
“We watch where the president goes,” says Dan Centinello, a Romney deputy political director who oversees the weekly meetings. “We’re trying to piece together what we think are his top ranks.”
That is easy enough, but Obama has not been traveling as a candidate for very long, so Centinello’s team doesn’t yet feel comfortable that they know how to interpret the logic of their opponent’s itinerary. Did Obama’s recent Ohio bus tour stop in Sandusky because his analysts detected a large pocket of Romney-leaning voters they thought they could win over? Or because they thought an Obama visit was necessary to jolt their own supporters’ enthusiasm levels? Or did Obama go to Sandusky because advisers wanted to draw a crowd to aid local efforts to register new voters or enlist volunteers?
In the primaries, Romney’s advisers had little confidence that there was much logic at all behind his rivals’ moves, and the two-time candidate outmaneuvered analytically amateurish opponents with well-plotted discipline and attention to detail. Now forced to play catch-up against a savvy incumbent, Romney’s team acknowledges they are not aiming to match what Obama has built in Chicago: A unique, in-house analytical empire that has developed an unrivaled capacity to churn through voter data and translate insights into tactics. Because of this capacity, Romney advisers assume that what they see the president doing in public must have a good deal of sense behind it. "The Obama team had the luxury of knowing exactly what they'd be doing on July 1, 2012 because they've been planning for six years—definitely three-and-a-half years,” says Zac Moffatt, Romney’s digital director. So instead of devoting their analytical energies to out-strategizing the president, Romney’s advisers are trying to hack Obama’s code and turn it against him.
As the dataset of Obama activity expands over the course of the campaign, the burden of finding those patterns will shift from the eyes of advisers huddled in weekly meetings to the statistical models they’ve written. Algorithms will test the association between vote goals and the candidates’ travel and ad placement, staring through Obama’s visible tactics to reveal a latent strategy beneath.
Those calculations, along with other data from internal polls and the campaign’s interactions with voters through field activity and phone contacts, feed into nightly simulations of local and state dynamics that spiral up all the way to electoral-college projections. Only then do Romney’s aides believe they will know enough about how their dollars ought to be spent, and where their candidate ought to go.
Throughout the primaries, competing as a well-known front-runner within a relatively small universe of potential voters, Mitt Romney gave his targeters one central task: identify his supporters so they could be mobilized to turn out. Romney demonstrated analytical acumen on the path to the nomination— his campaign’s ability to count votes in Iowa and savvy handling of the early-vote process in later states were both essential to his win—but he developed little institutional expertise along the way. Romney relied exclusively on his party’s leading data firm, TargetPoint Consulting, to build the microtargeting models that undergirded those tactics and to advise him on assembling the universes of individual voters to be contacted by mail or phone. When Obama clinched the 2008 nomination he had more than 10 times the 89 paid staff on Romney’s lean team when he extinguished Rick Santorum’s challenge this spring.
The turn toward a general election has expanded the Romney campaign’s analytical needs, while also refocusing them. The political department has instructed its field staff to be more aggressive in collecting information on people who attend Romney rallies so that targeters can build models to predict not only how a person will vote, but their likelihood of attending campaign events or agreeing to volunteer. While the political department is developing a plan to use those volunteers for get-out-the-vote activities this fall, the campaign’s targeters are more concerned with finding minds to change: those already likely to vote but still not sold on Romney. “The challenge is finding that share of persuadables that are out there,” says TargetPoint president Michael Meyers. “How do I find them and what do you talk to them about?”
When nominee Obama found himself in a similar position four years ago, his campaign decided to upend the standard structure campaigns used to interact with data. Obama had begun 2007 relying exclusively on an outside consultant, Ken Strasma, to develop his statistical models. But as the general election arrived, the campaign’s ambition for using data exceeded the reach of Strasma’s small firm. Expertise was no longer so centralized: by the end of the primaries dozens of employees had worked directly with Strasma’s microtargeting scores to build targeting universes and organizing maps, often improvising to solve local problems. That June, campaign manager David Plouffe approved a proposal to augment Strasma’s work by having staffers direct regional data and targeting desks at Obama headquarters. In Chicago, targeting and analytics would be treated as a core internal campaign function, like press relations or constituency outreach.
John McCain stuck to the old model, hiring outside consultants (largely through the Republican National Committee) who struggled throughout the summer to get the attention of campaign leadership. Those targeters had to wait until August to begin to building their state-specific statistical models that generate probability scores for individual voters. In nearly all of those states, McCain’s campaign never updated those scores, even as the dynamics of the race changed through such heady disruptions as Sarah Palin’s nomination and the collapse of Lehman Brothers.
While McCain’s campaign treated microtargeting as a one-time process, Obama made it nearly continuous. Targeting-desk staffers processed the results of hundreds of thousands of weekly voter interviews placed by Obama’s call centers. Starting that summer, they remodeled the electorate in every battleground state each weekend, so that when Palin arrived on the scene or the global-financial system collapsed field staff could see the events’ impact on the projected behaviors and beliefs of every voter nationwide.
That remains the blueprint for Obama’s 2012 data structure, with dozens of analysts spread among departments in Chicago—some assigned to specific regions and states and others available to work on special projects, such as the text-analytics initiative called Dreamcatcher—and no dominant outside analyst like Strasma under contract. Innovation to improve their strategy and tactics, Obama campaign officials have gambled, should come from within.
Even as Romney now reaches fundraising parity with Obama, his campaign isn’t aiming to duplicate Chicago’s analytics structure. A reliance on staff talent, Romney’s advisers caution, could potentially prevent a campaign from benefitting from ingenuity elsewhere. “The other side has certainly invested a lot of money,” says Meyers, “but they’ve been operating outside the market for two years.” Meyers’s TargetPoint, which works for a variety of campaigns and independent groups, including the pro-Romney super PACs Restore Our Future and American Crossroads, will handle the majority of states for Romney. The remainder have been assigned to Grassroots Targeting, its leading competitor among Republican data consultants.
Already Romney is moving more quickly than McCain’s did four years ago. His campaign began assigning its scores this June, and engineered systems so that new information from Republican field canvasses and phone banks can help to refine the statistical models on an ongoing basis. “It’s getting a lot faster and a lot more dynamic,” says Alex Lundry, a veteran of the McCain effort and perhaps the most methodologically sophisticated opinion researcher working in Republican politics.
Lundry recently moved to Boston to work for Romney full time, charged with leading a data-science team within the campaign’s strategy department. No longer competing in a few states at a time, strategists now need to make continuous determinations about how to move resources across a national map, and balance different tactical needs in each state. Lundry has been paired with Brent McGoldrick, who oversaw George W. Bush’s 2004 West Virginia operations but eventually left campaigns to join Financial Dynamics, a consultancy that works with clients in the health care, energy, and financial-services sectors to segment consumers by their attitudes and behaviors, and adjust corporate advertising strategies to better reach them. “Lots of us, including myself, have always done politics,” says Blaise Hazelwood, who launched Grassroots Targeting with McGoldrick in 2005. “He has had the time to experiment with all this. In politics you don’t typically have that type of opportunity—or not with the time or budget to do it.”
Romney’s digital department has gone on a summertime personnel spree, hiring former employees of Apple, Google Analytics, Ominture, and—in the case of the new manager of Romney’s online store—Overstock.com. (A cluster of newly hired engineers have been permanently situated in Utah, partially to exploit the time difference when working on overnight projects.) But the bulk of the online analytics come from commercially available services marketed by outside firms. Web ads are largely directed through segments packaged by the company Lotame based its own online-behavioral models. Other vendors have been enlisted for their boutique products, as ClickZ has reported: Pulpo for profiles of Hispanic sub-segments online, and Say Media for online video targeted at voters who do not watch much television and are hard to reach through traditional ads. "I don't think we thought, relative to the marketplace, we could be the best at data in-house all the time,” says Moffatt. “Our idea is to find the best firms to work with us."
It is a sentiment that aligns with the candidate’s own private-sector triumphalism, but more importantly reflects the Romney campaign’s acceptance of the David and Goliath dynamic between Boston and Chicago where analytics are concerned. Romney aides scoff at the glowing pieces written about Obama’s data-driven methods, but their obsession with reverse-engineering his analytics is its own concession. It is the statistician’s version of trailing a motorcade in a honking bus to find out where the president is headed, in the hopes of later divining why. “It’s one thing to know ours,” says Centinello, “but it’s even better to know what his strategy is.”
Sasha Issenberg is the author of The Victory Lab about the new science of political campaigns.