Chat with us, powered by LiveChat

How Far Can Political Technology Reach?

If you are a living, breathing, voting-age American, the political analytics firm i360 says it has your number. Its database contains more than 250 million individual profiles, basically one for every adult in the country, each of which is scored on a 100-point scale. Zero represents the most partisan Democrat, and 100 the most partisan Republican. The higher the score, the better it is for i360, which is affiliated with the conservative nonprofit network funded by the Koch brothers. The firm claims that its predictive model helps its clients to judge whether any given person will likely be receptive to an advertisement or a knock at the door — and will ultimately be inclined to vote the right way.

The company pinpoints a person’s place on its spectrum by using information from sources obvious and arcane: voter registries, credit agencies, social networks, and so on. Are you rich or poor? Do you live in a heavily Democratic precinct? Are you related to a lot of Republicans? Have you tweeted displeasure with President Obama? Did you recently purchase a Bible? Any of these data points might move the needle, i360 says.

Like most players in the world of digital politics, i360 maintains an air of mystery around the inner mechanics of its algorithms, and the firm keeps a low profile, rarely granting interviews. It recently moved into a new office in suburban Washington, which I was asked not to describe in any further detail. Although its president, Michael Palmer, declined to discuss specific clients, previous reports and public records indicate a tight relationship with the Koch-supported organization Freedom Partners and an affiliated nonprofit called the Themis Trust, which has spent millions to build an independent analytical capacity. “An outside entity can invest in the right things, with a long-term vision,” Palmer says. “A lot of the stuff that we’re building right now isn’t going to be ready until 2016.”

Of course, the basic idea behind i360 — the notion that technology can open a portal into voters’ minds — is nothing new. Political professionals, who work at the nexus of ideology and marketing, have been developing microtargeting techniques for years. And in 2012, Obama campaign staffers — working from a laboratory called “the Cave” — famously mined data to new extents of granular precision.

(Justin Sullivan/Getty Images)

What has changed since then — and continues to change, almost by the day — is the sheer number of firms and consultants operating in the political-technology sector. Huge sums of cash are flowing into it. Parties are fundraising for it. Billionaire venture capitalists are investing in it. Start-ups that specialize in it are constantly being born. With one eye on the midterms and the other on 2016, political technologists are constructing new weapons and experimenting with increasingly personalized targeting. And the competition between Democrats and Republicans is sure to intensify as they head into a wide-open presidential race, where billions of dollars will be spent under radically loosened campaign finance regulations.

Republicans, in particular, have been aggressively pressing to reverse-engineer the Obama campaign’s tactics. “They iterated forward, but I don’t think the vision has been fully realized yet,” Chuck DeFeo, the Republican National Committee’s chief digital officer, said in a presentation at a recent technology conference called the Personal Democracy Forum. He described an R&D effort that includes a Silicon Valley “incubator,” run by an engineer recruited from Facebook, that is reportedly at work on a computer system called Beacon. But as is often the case with advanced technology, the specifics remain difficult for outsiders to evaluate. “For the first time, I’m going to publicly show what that platform looks like,” DeFeo promised the audience, but he was just joking. When he flipped to his next slide, the giant screen behind him projected the menacing image of the Death Star.

For all the futuristic rhetoric that surrounds the growing role of technology in politics, the actual capabilities of the new tools aren’t well understood — not by the general public, and not even by many people who work in politics. “There’s a lot of snake-oil salesmen coming through, taking advantage of the lack of knowledge,” says Zac Moffatt, the digital director for Mitt Romney’s 2012 campaign who now runs the firm Targeted Victory. Politics, advertising, and the tech business are all prone to exaggeration, and when they come together, the hucksterism can get grandiose. Poke around the subject, and your email in-box will soon begin to fill with subject lines that read like ads for male enhancement: “Deep Root Analytics: 5 extra points at the polls.”

Can any technology really deliver those kinds of results? And if so, how does it actually work? The digital sphere of politics encompasses a wide variety of functions, including data-gathering, analysis, and ad technology — and some of them are better developed than others. There’s no doubt that there’s been an explosion in the amount of data available to campaigns, as every variety of human interaction — conversation, commerce, friendship, entertainment — moves into the incessantly monitored and quantified online realm. And it’s equally clear that the Obama campaign’s innovations will leave a legacy as the volley that set off an arms race. But as potent as their surveillance may be, the technologists still face a challenge: How well can they translate everything they know about voters into votes?

“There’s a lot of snake-oil salesmen coming through, taking advantage of the lack of knowledge.”

THE INNOVATORS ARE always working around a central irony: The very advances that make it possible to know so much about voters also make them more difficult to reach. A DVR records your viewing habits, but it also allows you to fast-forward through the standard 30-second campaign spot. Spam filters are rising; network audience numbers are falling. It takes plenty of invention just to counteract the relentless force of media entropy.

The process of finding and persuading a potential voter progresses in stages; consumer marketers, who use similar methods, employ the metaphor of the “purchase funnel.” At the top of the funnel sit firms that gather and analyze data, like the Democratic powerhouses Catalist and NGP VAN, or i360, which is trying to reproduce the model for Republicans. The analytics firms, in effect, describe voters to campaigns, helping them to size up the marketplace for their messages. Although the methods they employ are alarming to some privacy advocates, they are perfectly legal, and both parties use them avidly.

Like Google, the National Security Agency, or the Democratic data machine, i360 has a voracious appetite for personal information. It is constantly ingesting new data into its targeting systems, which predict not only partisan identification but also sentiments about issues such as abortion, taxes, and health care. When I visited the i360 office, an employee gave me a demonstration, zooming in on a map to focus on a particular 66-year-old high school teacher who lives in an apartment complex in Alexandria, Virginia. If a campaign wanted to reach her, it could send a volunteer to record her opinions on i360’s smartphone app, while tracking the encounter via GPS. If she seemed receptive, the campaign could use what consumer marketers call “look-alike modeling” to locate others with similar profiles. Though the advertising industry typically eschews addressing any single individual — it’s not just invasive, it’s also inefficient — it is becoming commonplace to target extremely narrow audiences. So the schoolteacher, along with a few look-alikes, might see a tailored ad the next time she clicks on YouTube.

“We say every different demographic and constituency can be reached through different means, at different times in the election, but here’s the people you need to talk to,” Palmer says. “Is this a perfect science? Do I know exactly what is going to move you? No, but I have a higher degree of probability of moving you using this data.”

{{thirdPartyEmbed type:magazineAd source:magazine_mid}}

The influence of analytics has risen, in part, to compensate for the increasingly glaring failings of traditional telephone polls. About a third of households use only cell phones, making them tough to call, and poll response rates on landlines have fallen from about 40 percent to less than 10 percent since the 1990s. Small samples require much statistical tweaking to yield accurate results. “Every single cycle it gets more and more difficult,” says David Rothschild, an economist who has studied polling for Microsoft Research. “As response rates fall further and further, it becomes more likely that they may miss that critical point where the sample is no longer more or less representative. And once that happens, it is not going to work anymore.” 

The less representative the sample, the greater room for error — or even wishful thinking — to creep in. Romney’s team, to its subsequent humiliation, believed internal polls that showed him running ahead in several battleground states on the eve of the 2012 election. The Obama campaign, by contrast, attempted to fill the knowledge gap with analytics. Many capabilities that firms like i360 now tout were developed by the president’s team in 2012, which built on advances the Democrats made in 2008, which in turn drew from the microtargeting success of George W. Bush in 2004. Facing a tough reelection challenge, Obama’s campaign strove to understand the electorate “at an atomic level,” the journalist Sasha Issenberg recounts in his book The Victory Lab. “Obama’s campaign began the election year confident it knew the name of every one of the 69,456,897 Americans who had put him in the White House,” Issenberg writes. “They would reassemble the coalition, one by one, through personal contacts.”

“We say every different demographic and constituency can be reached through different means, at different times in the election, but here’s the people you need to talk to.”

To carry out that mission, Obama put together a huge ground operation, backed by an analytics staff of more than 50 people — 10 times the size of Romney’s — and a team of embedded social scientists that tested the persuasive power of the campaign’s message. Instead of just relying on telephone polls of supposedly likely voters, it produced reports that attempted to gauge the sentiments and demographic makeup of the electorate based on hundreds of data points. “The theme is that there are more diverse means of measurement,” says Dan Wagner, an Obama campaign number-cruncher who, with funding from Google’s Eric Schmidt, subsequently founded a start-up called Civis Analytics. “The thing about online is, you can collect data very cheaply, very widely, and very quickly.”

At least in theory, political campaigns start with a good idea of who might be buying their product. Voter registrations are public record, and both parties maintain voluminous files of names. But as a practical matter, registration information is dispersed haphazardly across 50 states and must be constantly refreshed. And even then, it can only reveal so much about how someone is going to vote.

That’s where analytics can help campaigns figure out the electorate. Crude techniques have existed for years: filtering the voter file according to demographics such as age or race, or using historical precinct results, or cross-referencing the membership lists of organizations like Planned Parenthood and the NRA. At a more sophisticated level, campaigns can draw on reservoirs of individual consumer data that agencies such as Experian use to compile credit scores. There are ways to bore in on specific types of purchases. But indicators such as income level and gun ownership are still just proxies. It was only recently that analysts began to be able to be tap into a much more reliable gauge of public opinion: what people reveal about themselves when they imagine they are just among friends.

(Jewel Samad/Getty Images)”The group of people on Facebook are not representative of the country yet, but they are representative of all those people on Facebook,” Wagner says. What makes a site like Facebook valuable to advertisers is the information it continually gathers, even when you don’t think it’s looking. “The thing about the new social networks is that they are creating unprecedented consumer profiles,” says Max Kalehoff, senior vice president at SocialCode, a commercial ad tech company. “Most people are logged onto Facebook whenever they do anything else on the Web. It creates this incredible bread-crumb trail.”

Many of the basic techniques used by Obama’s analysts were originally developed in the private sector — by ad tech companies that silently record and categorize online behavior, reducing individual consumers to sets of data points. Ad tech has advanced with great leaps in recent years, bringing the industry close to its omniscient goal: being able to reliably pinpoint what a single person is up to on his or her computer, TV, and phone, all the time. Mobile technology, in particular, has opened many new monitoring opportunities. If you have a social-networking app on your phone, it may know all sorts of things about your daily interactions, whether virtual or literal. “It connects everybody through those platforms pretty much 24 hours a day,” Wagner says. “The device is almost irrelevant. What’s valuable is the network.” 

While the amount of personal information available to campaigns keeps expanding, however, the limits of the public’s tolerance for its use are still being tested. Facebook recently ran into controversy when it was revealed that its Data Science division was experimenting with instilling positive and negative moods in users through their news feeds. Politicians are understandably wary of looking like spies. A 2012 University of Pennsylvania study found that 64 percent of Americans would be less likely to vote for a candidate if they knew he or she was buying data about their online behavior.

{{thirdPartyEmbed type:magazineAd source:magazine_bottom}}

Yet even as people complain about intrusion, any user of Facebook or Twitter knows there is little limit to what many will share about their lives and opinions. Rothschild, the Microsoft researcher, has been experimenting with ways to put this connectivity to use, through polling systems that rely on responses to surveys on devices like the Xbox. In a forthcoming paper, authored in cooperation with scholars at Columbia University, Rothschild writes that even an unrepresentative population such as video-game players could be made to yield statistically reliable results if the sample size was large enough. “I think you’ll see an explosion of this type of polling in 2016,” says the economist, who has discussed his work with operatives on both sides of the aisle.

However, Rothschild says, there’s a practical question that remains unanswered: What’s the value of all this information? “We’re doing a good job at collecting data and a decent job at analyzing it,” he argues. “But we’re doing a really mediocre job at using that data.”

ONCE A CAMPAIGN has compiled its list of targets, there are still only a few ways to deliver the political payload, and none of them are really new. The first is through a mailbox, either postal or digital. Political consultants have built a lucrative industry around direct mail, but its modern impact is debatable. Romney overwhelmingly outspent Obama on direct mail. Meanwhile, Bill de Blasio won last year’s competitive New York City mayoral primary without sending out a single mass mailing.

The second way to deliver messages is through personal interactions. Voters can be convinced by volunteers at their doors or friends in the grocery aisle — or on Facebook. Traditionally, door-to-door canvassing has been blue-collar work. But technology has upended that hierarchy by giving a futuristic sheen to grassroots mobilization. Social networks make it possible to coordinate personal interactions on a massive scale. Jim Gilliam, founder of NationBuilder, a nonpartisan organizing platform, says his technology allows campaigns to follow what people are saying online, so they can ultimately enlist core supporters to recruit their friends. “It all flows into one database,” he says. “Basically, what’s going to happen is that anybody who expresses interest in the most modest way can be followed up with by someone who cares.”

“TV is ripe for disruption. There are so many inefficiencies to exploit.”

NationBuilder is funded by millions of dollars in venture capital from tech investors such as Marc Andreessen and Pierre Omidyar. Gilliam has a background in progressive causes, but the company does business with any client that is willing to pay, which makes NationBuilder both a rarity in political circles and an object of some suspicion. (The company’s agnosticism has also made it a target for hackers, who recently brought down its network with what Gilliam described as an attack “directed at one of our customers for their political beliefs” and “meant to disrupt upcoming elections.” I was told the customer was Britain’s anti-immigrant party UKIP.) Gilliam rejects the notion that programming should be tied up in partisan politics. “It’s just not possible to get that technology solid in the high-pressure environment of a campaign, where you basically don’t care if it works after Election Day,” he says. “That’s why we’re structured as a tech company.”

Indeed, political campaigns form and disband quickly, and infighting is rife inside party committees, with millions in campaign spending at stake. Recently, The Huffington Post reported that Andy Barkett, the chief technology officer the RNC hired away from Facebook with much fanfare, had already been sidelined amid dissension. Even the vaunted Obama campaign had its share of internal turf battles — and struggled at times to translate the analytics performed in the Cave into on-the-ground action. “It’s a little bit the untold story of the Obama campaign,” Gilliam says, “because it’s just not as sexy as, ‘There was this gigantic computer database that was collecting votes.’ “

One of the Obama campaign’s most-hyped innovations was its use of “targeted sharing” — basically, a form of friend-to-friend mobilization ginned up through a customized Facebook app. The campaign boasted that it had generated 5 million contacts through targeted sharing in 2012, and afterward it became a popular buzzword in consulting circles. But Moffatt, the former Romney strategist, points out that a pair of academics later produced an estimate that the program may have resulted in a mere 20,000 votes for the president. “Postelection, there’s always the shiny object,” Moffatt says. “Conceptually it sounds so cool that people are willing to pay astronomical amounts for it.”

Andy Barkett speaks during the Republican Jewish Coalition spring leadership meeting. (Ethan Miller/Getty Images)USING SOCIAL NETWORKS ;to muster volunteer armies sounds great at conferences, but, in practice, political operatives know you tend to get what you pay for. Ultimately, if a campaign wants to make its case, it needs to advertise. For most large campaigns, the paid media budget is the largest area of expenditure, and typically around 90 percent of it is spent on TV. But an untold proportion of that money is wasted. A recent survey conducted for Google and several political consulting firms found that over the course of the prior week, one-third of likely voters had watched no television programming at the time it was broadcast, outside of sports.

That still leaves two-thirds of the electorate that can be reached through television, but they are now spread across hundreds of available channels, which charge advertisers widely varying rates for airtime. In 2012, the Obama campaign tried to figure out how to spend its television budget more efficiently, developing something called the Optimizer. This system cross-referenced the names of campaign targets with data compiled by a company named Rentrak, which has access to millions of cable boxes. Instead of simply buying ads on CNN and the nightly news, like most candidates do, the Obama campaign figured out what shows its desired audience was watching, and how to reach them at the most cost-effective time slots.

“The Romney campaign outspent us 8-to-1 on the news because they were working from the old playbook,” says Chauncey McLean, who worked on the Obama campaign and, after the election, cofounded a company called the Analytics Media Group to refine the Optimizer’s technology and market it to political campaigns and commercial brands. When I visited McLean’s New York City office recently, he showed me a report prepared for a candidate in the San Diego television market, which recommended buying late-night reruns of shows like Leverage and Family Guy. Since 2012, the tactic has become commonplace, and Republican firms such as i360 have formed similar partnerships with Rentrak.

“TV is ripe for disruption,” says McLean’s partner Chris Frommann. “There are so many inefficiencies to exploit.” Just over the horizon, they say, is “addressable” TV advertising, which is already being tested on some cable systems. By 2016, it’s possible that you and your neighbor may be watching the same show but receiving different commercials based on your demographics. “If you have your target audience down to the individual level, it unlocks some very powerful marketing tools,” McLean says.

(Photo Illustration by Jenny Mazer. Images: Adam Berry/Getty Images, left; Andy Jacobsohn, right)The techniques that AMG uses to place TV commercials more efficiently are adapted from digital advertising, where targeting can be performed with far more precision. In recent years, as consumer activity has moved online, many commercial brands have shifted a third or more of their marketing dollars to digital media. But politics has been slow to follow suit. The Obama campaign was comparatively aggressive, spending about 20 percent of its $580 million advertising budget online. Jim Walsh of the Democratic ad-tech firm DSPolitical predicts that the trend will escalate in 2016, as campaigns seek to turn all the knowledge they are gathering on individual voter preferences into customized messages. “The amount of dollars going into digital advertising for targeting,” he says, “is going to create a massive war.” 

DSPolitical bills itself as the “home of the political cookie.” As most savvy computer users know, cookies are unobtrusive homing beacons that your Web browser picks up as it traverses the Internet. If you recently looked at a pair of shoes online and now see shoe ads everywhere, you’ve seen the technology at work. Walsh says DSPolitical maintains a pool of around 600 million cookies and mobile-device profiles. If one matches up to a political client’s list of targets, it can be served an ad. Walsh likens the effect of Web advertising to the billboards that tout South of the Border, a tourist trap on I-95 in the Carolinas. “For 100 miles, imagine you, your wife, and your two kids in the backseat, all seeing the same billboard every 15 feet,” he says, “but for each of you in the car you’re seeing a totally different message that’s literally targeting that individual person.” You may not notice the message the first time, or the second, but eventually it penetrates.

This form of advertising is known as direct marketing: It is meant to inspire a specific action, such as pulling off the interstate to buy cheap fireworks, or clicking on a link to donate to a candidate. Direct marketing has traditionally been regarded as a dark alley off of Madison Avenue — a grungy counterpart to the more lucrative and glamorous “brand” advertising that viewers are accustomed to seeing on television. Brand advertising is designed to make you think, “Obama: Hope” or “Chevy trucks: They’re like a rock,” and it has a nostalgic television series about its glory days. But direct marketing pretty much owns the Internet. Google makes its billions by hosting tiny little ads for lawn mowers to people at the very moment they are searching for a lawn mower.

Digital advertising firms use a similarly direct approach to market candidates and ideologies. “If you’re not using this solution,” says Rich Masterson, founder of the Republican firm Campaign Grid, “then 90 percent of your ad money is wasted.”

The placement of Web advertising is typically done via ad networks that sell massive numbers of Web impressions across many sites. One morning in May, shortly before this year’s primaries in Pennsylvania, I visited Campaign Grid’s suburban Philadelphia offices, where Masterson gestured across a room of workers in cubicles. “It’s just like a Wall Street trading floor,” he said. “Except instead of trading pork bellies, they’re trading audiences for 105 campaigns.” 

Masterson doesn’t have a political background. His formative experience is in digital marketing. He started a company called U.S. Interactive that rose and crashed during the first dot-com boom. In 2008, though, he was thinking about running for Congress in Pennsylvania, and he discovered how campaigns typically operate. “The political consultant told me I had to spend $5 million on advertising in three different states,” Masterson says. 

A campaign could potentially track whether its targeted voters had visited polling places.

He looked for Web-based alternatives but found a void. “I just said, this is absolute insanity,” he recalls. “I was really shocked that nobody had solved it, because you’re talking about billions of dollars in political spending.” So instead of running for office, Masterson got together with a friend named Jeff Dittus, whose background was in the direct marketing of products like audiobooks and Thighmasters, and started Campaign Grid. Soon, they were working for candidates such as Chris Christie as well as the RNC, which Masterson says shared data that the company used to build its own proprietary voter file. Campaign Grid recently received a patent for what Dittus calls the “holy grail,” a system that purports to integrate the voter data with ad-buying programs to automatically place Web videos and banner ads in front of the right voters.

Because so much inventory is available across the Web, digital ads are dirt cheap, but measuring their effectiveness is a tricky business. Those on the creative side of advertising tend to be dismissive of generic banners, and their “click here!” crudity. “There is no doubt that technology has brought addressable messaging to new levels of sophistication,” says Andrew Essex, vice chairman of Droga5, a digital agency that specializes in brand advertising. “But if you’re still serving them something that is the equivalent of a tedious bumper sticker, it is going to be discarded and ignored.” 

Droga5, which works primarily with companies such as Kraft and Coke but also occasionally does political work, recently made a viral video in which celebrities’ mothers implored their kids to sign up for Obamacare. It’s been viewed more than 800,000 times. By contrast, a Web ad buy of 2 million impressions that Campaign Grid targeted to suburban women on behalf of a candidate in Nevada resulted in just over 1,000 clicks — a rate of .05 percent. That minuscule ratio is actually pretty much in line with industry norms. 

Maryland voters in 2012. (Chip Somodevilla/Getty Images)Even those numbers may not always be what they seem, however, due to the impenetrable and sometimes seedy economy of online ad networks. The audience research firm comScore has estimated that 36 percent of all Web traffic is generated by botnets, raising the possibility that many ad impressions are fraudulent. Investigations by The New York Times and The Wall Street Journal have revealed that video ads sometimes end up unwittingly displayed next to porn, or on unscrupulous websites that show multiple videos at the size of a single pixel. Several digital campaign firms betrayed their alarm by publishing written replies to the reports, arguing that the graft merely demonstrated the need for expert guidance.

Digital advertising firms argue that measuring clicks alone understates the effect of their marketing messages, because a response isn’t always immediate. Some are experimenting with other metrics, such as tracking the amount of time your mouse spends hovering over an ad. And even a tiny click-through rate can still generate a good return on investment if you consider the many millions of impressions that can be had for the cost of a single TV spot. A few years ago, Campaign Grid ran an attack on a Democratic House member who had voted to fund a study that tested the effects of cocaine on monkeys. The ad — which featured a screaming chimp and the slogan “Does Your Monkey Need Rehab?” — was deployed in some 14 million paid impressions over eight days just before the election. Afterward, a poll found that 4 percent of the people who managed to recall the monkey message switched their vote to the Republican.

These are marginal effects, but, like many ad tech salesmen, Masterson argues that the industry has only begun to explore the frontiers of direct marketing. “This device is pinging all the time,” he says, holding up his cell phone. “It’s a passive check-in.” Consumer advertising firms are currently developing systems that use GPS telemetry to “geotarget” consumers. A store can send you a coupon as you’re walking by to lure you inside, or a car company can advertise to everyone who walks into one of its competitor’s showrooms. Masterson foresees all sorts of potential geotargeting applications in politics. For instance, a campaign might track how many of its targeted voters had actually visited their polling places on Election Day.

“Political advertising is like building a brand based on somebody that you hate.”

VOTER REGISTRATION data can tell a campaign where you live, but mobile technology has the potential to reveal how you live: where you go, what you do, with whom you interact. Still, there are limits to technology’s power over the electorate. “It’s fine to understand whether someone is up for grabs,” says Chauncey McLean. “But can anything I say to you change your mind? Because, if not, it doesn’t matter.”

It does not seem like a historical coincidence that political tactics have taken on a microscopic focus at a moment when both parties are mired in alienating smallness. Targeting technologies tend to be most useful for motivating the converted, aiding the laborious work of organization and turnout. When it comes to the more nebulous component of politics — persuading the undecided — their effectiveness is less clear. Despite a century’s worth of psychological study, persuasion remains something of a mystery, even to the experts. “The stuff I am extremely skeptical of is this idea that we can turn data into ad campaigns and magically turn people into voters,” says Jim Gilliam. “That’s not real.”

Jim Ferguson, a long-haired Texan, is a veteran ad man who worked at firms like Young & Rubicam, where he came up with memorable branding such as “Beef: It’s What’s for Dinner.” In 2012, he joined Mitt Romney’s campaign as its creative director. “I came from a traditional brand-building background,” Ferguson says, “and one of the things we were always taught was, you buy from someone that you like.” But he discovered that the philosophy of persuasion in today’s campaigns is different. “Political advertising,” Ferguson says, “is like building a brand based on somebody that you hate.”

Mitt Romney greets workers at call center on November 6, 2012 (Justin Sullivan / Getty)Ferguson told me that the Romney campaign grew so preoccupied with targeting narrow slivers of the electorate that it never projected a broadly appealing brand identity. The campaign concentrated on making ads meant to directly appeal to base voters, such as the one that deployed an out-of-context line — “You didn’t build that” — from an Obama speech on infrastructure, an attack Ferguson now deems “such bullshit.” The ads motivated the faithful, but they didn’t change enough minds to win the election.

“All the evolved tools and innovations in political communications are fabulous,” says Republican media adviser Mark McKinnon, “but they don’t mean squat unless you’ve got a great message and a talented messenger.”

I figured that if anyone is really capable of using precision technology to persuade voters, it is the team that worked for Obama. So one spring morning, I attended a reunion of sorts at the Manhattan office of Bully Pulpit Interactive. Some 2012 campaign alums were now with the firm, which handled Obama’s digital media, and others had gone to work for Google, on the sale side of advertising. The meeting’s atmosphere was chummy, but the business at hand was potentially dismaying: the coming midterms.

As the group met around a converted ping-pong table, Andrew Bleeker, Bully Pulpit’s president, went over a long list of endangered congressional seats. “I can’t believe Colorado is going to be a thing,” said Jason Rosenbaum, a bearded Google executive who formerly worked for the Democratic Senatorial Campaign Committee. But more battlegrounds meant buying more advertisements via Google. In addition to working for Democratic campaigns, Bully Pulpit is designing advertising for big-spending independent committees like NextGen Climate Action, the liberal billionaire Tom Steyer’s super PAC. The firm told Google’s team that it planned to make heavy use of targeting to turn out voters. “Our buys, except in a very few cases, will not be national,” Bleeker said.

Mark Skidmore, a Bully Pulpit partner wearing a blazer with a pocket square and his hair in a faux hawk, asked about geotargeting phones and scraping server data to track opposition ads. Bully Pulpit is working to develop proprietary technology to monitor whether Web ads are actually seen, as opposed to just clicked on, in order to better test the effects of political messages. “This cycle, I’m very bullish on testing,” Skidmore said, “looking ahead to 2016.” 

Over lunch after the Google meeting, Skidmore, who used to manage marketing campaigns for Nike, said the Washington-based Bully Pulpit had opened the New York office to be closer to the cutting edge of ad tech. The company was particularly proud to show off some recent work it did for Steyer’s super PAC, which has announced it plans to spend $100 million to support Democrats and to press its environmentally centered agenda. “The organization is introducing itself,” Ben Coffey Clark, the Bully Pulpit partner who handled the campaign, told me as he prepared for a conference call with NextGen, where he would show the results of an initial campaign. “What you’ll see is a small part of a very large operation on the persuasion side, where we were introducing NextGen to elites.”

During the call, Clark said the campaign, timed to coincide with the White House Correspondents’ Dinner, had resulted in more than 11 million ad impressions. It included targeted page “takeovers” on Politico and The Washington Post and mobile ads that specifically hit users geolocated on Capitol Hill, on K Street, and at the Washington Hilton. Clark directed the meeting’s attention to one of the takeovers, which was meant to cast unflattering attention on Republican Sen. Marco Rubio’s support for the Keystone XL oil pipeline. The ad featured animation of Rubio’s head splitting in two to reveal the slogan: “Senator Rubio: Don’t Get Taken for a Sucker.” “The head splitting open was very much done in the line of high-impact poster art,” Clark said. “That real eye-catching font from the propaganda artists of the ‘60s.”

The image was striking, yes, but persuasion remains more art than science. If you were already convinced that the Keystone pipeline was an environmental outrage, the ad might inspire you to make a call to Congress, or send a Tweet, or turn out to vote for a Democrat. If you were unsure of the merits, though, would an evocative attack on Rubio change your mind? Convincing the uncommitted isn’t as easy as flipping a switch. That, perhaps thankfully, remains beyond the grasp of technology. Politicians still must contend with an old-fashioned problem: It’s not enough to get inside voters’ heads. You have to have something to say once you’re there.

Andrew Rice is a contributing editor at New York Magazine.

This article has been updated to include a longer version of Jim Walsh’s original quote about seeing the same billboard every 15 feet.

Sign up for our Newsletter