The Victory Lab (28 page)

When Kaine won, largely by making significant inroads into exurbs that Bush had carried, it was heralded as a harbinger of a Democratic comeback. Even though their scores had been developed specifically for Kaine’s race, Quinn and Steitz thought that the numbers could help others running lefty campaigns in Virginia—from a candidate for the state house of delegates to the League of Conservation Voters—in the same way that credit scores were useful to anyone looking to make a business decision. After Bush’s reelection,
microtargeting
had become a faddish buzzword, but it remained inaccessible to most people in politics. Large commercial data warehouses like InfoUSA, Acxiom, and Experian were all trying to sell their products to political campaigns, but their products were priced for corporate clients. Typically they charged hefty rates for every record
pulled from the master file; while presidential reelection committees and national parties could afford such costly data bills, they were out of reach for local campaigns and advocacy groups. “One of the things that pricing model does on the political side is it creates incentives for people to use as little data as possible and as late as possible,” says Quinn. “The idea was deliberately to change those incentives, to give people more data sooner, so they could experiment, consistently measure themselves.”

Even so, many smaller groups would also not know what to do with such a large array of data when they got it. A single category of raw data had obvious applications for commercial marketers; when Home Depot wanted to know where to put a new store, it made sense to buy a list of people in a given ZIP code who had swimming pools. But political campaigns were less interested in a single consumer variable than in combining many of them with election-specific information, such as polls or phone bank IDs, to find patterns in the two sets. That took a level of expertise and computing power that didn’t exist in the average statewide campaign headquarters or union hall.

In April 2006, Quinn and former Clinton White House adviser Harold Ickes folded Copernicus into a new company, Catalist, that would serve what they described as a data “utility” for Democratic campaigns and liberal causes. Ickes, who had helped to found America Coming Together, had returned to many of the same donors who had funded that effort. Placing calls from a hospital bed as he recuperated from a shattered hip incurred in a Vespa accident on Pennsylvania Avenue, Ickes made the case that the left needed a unified data infrastructure if it was ever going to catch up with what Republicans had built to prepare for Bush’s reelection. “Two thousand four really got big donors and movers and shakers on the progressive side interested in voter contact. Philosophically they kind of liked it,” says Ickes, who won the first $1 million commitment toward his $5 million goal from financier George Soros. “George comes from the European mentality and I think he was very enthusiastic about voter contact, much more than media—he didn’t give the back of his hand to media but
thought voter contact was something that we ought to, as progressives, get involved in.”

Ickes was asking would-be backers like Soros to be investors rather than donors, but in a company that declared its “double bottom line”: Catalist would pay attention to revenue but was more interested in keeping its prices down to help partisan and ideological allies win elections. Incorporating as a business would, in an era of restricted campaign finance laws, allow Catalist to operate as a for-profit vendor and not a nonprofit committee (as ACT and others were). They could do business with candidate campaigns, parties, and outside groups like unions all at once, as long as the flat, unlimited-access fees they charged each were considered fair-market value. In the same way that a video archive could legally license the same waving-flag footage to John Kerry and the AFL-CIO’s independent-expenditure committee without being accused of coordinating between the two, each group could be a Catalist client. They could buy the same voter files, with the same modeling scores attached.

When clients signed on, Catalist account representatives would ask for old voter IDs compiled from canvassers and phone banks. The best-case scenario was often being introduced to a desk drawer filled with records from past elections that the groups had never synthesized into a permanent file. In Catalist’s hands, each individual ID would be another data point that the algorithms could use as they profiled individuals. Catalist’s “ballots cast” table—the simple voter-file category of who voted in which election—quickly had well over a billion pieces of data in it alone.

Already, Catalist was operating on an astounding scale. Because many Catalist clients wanted data that could help them target nonvoters in registration drives, the company aspired to have a record for every voting-age adult in the United States, unlike the party’s databases, which included only registered voters. The standard statistical-processing software used for those who had done microtargeting on a state-by-state level, or had managed a single union’s membership list, couldn’t handle much more than ten million records. Catalist was building a file with 200 million—with the
exception of the credit-rating agencies, very few other private-sector entities ever needed to look at the whole country at once. “The things happening in the political sphere were approaching the more sophisticated techniques being used” in the corporate world, says Vijay Ravindran, a former
Amazon.com
software developer who became Catalist’s first chief technology officer. Catalist now maintains one-half of a petabyte of data, the equivalent of one thousand hard drives.

In Catalist’s early days, as Ickes tried to line up investors and Quinn recruited clients among liberal interest groups, they felt they had to fend off what Ickes describes as “backbiting” from party leaders he suspected of maneuvering behind the scenes to discourage support for Catalist. At the Democratic National Committee, Dean was working to complete the costly national voter file project that McAuliffe had started. Party officials made the case that Catalist would be redundant, but Ickes believed that they simply sensed the threat from a new power base that would sit outside the DNC, part of an inexorable shift in influence from the parties to independent players. “Dean had the aura of being the tech candidate, and knowing everything about technology,” says Ickes. “I think the establishment just didn’t want something that the establishment couldn’t control.”

Quinn and Ickes weren’t the only Democratic operatives trying to build common political tools that could be shared across the left.
Throughout 2004, a Democratic consultant named Rob Stein had traveled the country meeting the wealthiest liberals he could find to show them a series of PowerPoint slides he had prepared. Democrats had long agonized over what they saw as the right’s supremacy in developing political soft power: the web of well-funded conservative think tanks and foundations, generating ideas and arguments that flow to voters through sympathetic media channels. Stein’s presentation, “The Conservative Message Machine’s Money Matrix,” validated the existence of what Hillary Clinton in 1998 had called a “vast right-wing conspiracy,” only Stein did it with an accountant’s authority. He assessed the matrix as a $300 million annual operation, and showed how the right systematically outspent the left in every
category of political communication: $170 million to $85 million in think tanks, $35 million to $5 million in legal advocacy groups. Stein had begun his research after the 2002 midterms, but after Kerry’s loss his argument burned with new urgency. Beginning in late 2004, Stein persuaded Soros and Peter Lewis—
who had each given $20 million to launch America Coming Together and its affiliated Media Fund—to look past short-term organizing and toward long-term institution-building. Soros and Lewis led other liberal donors to band together as the Democracy Alliance, a loose confederation that would direct investment into new projects. The group, and its participants, directed money to keystone institutions of the new new left, including the Center for American Progress, which rapidly emerged as one of Washington’s most influential think tanks, and Media Matters for America, a watchdog group that worked, largely in vain, to embarrass the Fox News Channel and Rush Limbaugh.

At the same time, Kerry’s defeat also quietly inspired an era of entrepreneurship centered on the hard power of winning votes. The New Organizing Institute was launched in 2005 by young veterans of the Kerry and Dean campaigns to train operatives in online organizing. The Atlas Project established an archive of old strategy plans and polling memos from Democratic campaigns, the type of thing generally lost to trash collectors the morning after the votes are counted. “All of this stuff together is building an infrastructure for Democratic candidates and progressive groups to change the way business is being done,” says Steve Rosenthal, the former AFL-CIO political director who helped found the Atlas Project.

Meanwhile, Podhorzer’s geek lunches continued, and even though he was not actively recruiting participants the community grew slowly by word of mouth. The biweekly sessions became known as the Analyst Group, and over sandwiches and diet Coke and a big green salad that often went untouched, many of the new attendees ended up revealing the tinkering they had been quietly doing within their organizations. “One of the things about the Analyst Group that’s been jaw-dropping is how much people
report failure,” says Hesla. “It’s like AA for geeks: it’s this very trusting environment where everyone shares.”

By the time the Democrats won back Congress in late 2006, Podhorzer had moved the lunch discussions to the President’s Room on the first floor of the AFL’s headquarters to accommodate the sixty people regularly showing up for them. The group’s reach, in fact, was becoming so broad that it needed a structure to coordinate new research projects and share findings. In 2007, Podhorzer and his circle created the Analyst Institute, designed to operate with the sensibility of a think tank but the privacy of a for-profit consulting firm.

Quinn and Steitz joined the board, and they had an idea that the new entity’s destiny had to be interlocked with Catalist, each a political business founded with little interest in profit. Steitz thought back to Nigel Morris, whom Steitz had approached in late 2005 to discuss how American progressives could adopt lessons from the testing culture of Capital One. If the company wanted to stay ahead permanently, Morris said, it could do so only by perennially experimenting—testing everything, and then agglomerating the data in a way that could reveal new insights about each individual in their database over time.
Already the company ran nearly fifty thousand experiments yearly, part of a goal to have more “laps around the track” than any competitor, as Morris put it. Steitz began to adopt the metaphor as his own. Catalist and the Analyst Institute were, he thought, the vehicles the left had long needed. “The progressive movement,” Steitz wrote in a memo, “needs to increase our laps around the track by orders of magnitude.”

Podhorzer became the group’s chairman and cast about to hire an executive director who would decide what the left’s next round of experiments would be. Gerber and Green recommended one of their protégés, David Nickerson, who was then teaching at Notre Dame and running his own battery of experiments. But Nickerson turned down the job, worried that he would be performing “proprietary science” for activists rather than making his work public in scholarly spaces.

Gerber then suggested Todd Rogers, on whose dissertation committee he had recently served. Rogers had an unusual profile to run a Washington consulting firm: a psychologist who graduated from Harvard Business School after performing research that examined whether the way individuals managed their Netflix queues could illuminate how they felt about a carbon tax to fight global warming. Elsewhere in his dissertation, Rogers had run an experiment that fooled citizens into visualizing themselves going to the polls, and then measured whether they actually did. Another proposed that it might be possible to drive up turnout merely by talking it up as a popular activity.

These were turnout levers that the Analyst Group members had never before thought about pulling. Green said he was struck by Rogers, not yet thirty, as “a throwback to older days in which psychologists were interested in the dynamics of social action.” Podhorzer was attached to a more modest historical analogy. The statistical revolution that had fascinated a teenaged Podhorzer had transformed baseball, as many of those who had grown up using numbers to undermine the wisdom of elders now ran their own front offices. In an early job posting after he moved to Washington as the group’s director, Rogers wrote, “Some people have described what we are trying to do as ‘
Moneyball
for progressive politics.’ ”

WHEN SHAME PAYS A HOUSE CALL

I
n 2002, Todd Rogers opened a manila file folder and let a flurry of newspaper and magazine clips settle on the kitchen table of his Western Massachusetts home. Less than two years earlier, the recent Williams College graduate had turned down opportunities to work at bigger Washington polling firms to join Abacus Associates, whose small office was run by two former academics and where Rogers handled junior tasks like data analysis and writing survey questions. But now he was itching to do something new, as was made clear by the fact that he was opening the folder. When Rogers started at Abacus, he had christened the file, and gradually filled it with news articles: every time he found himself intrigued by something outside his intellectual comfort zone, he would stuff it in the folder. As he spread years of clips on his table, he was trusting them to guide him toward a new career. “Eventually it all made sense. It was all about behavioral science experiments,” says Rogers. “Your interests reveal themselves.”