Big Data Is Watching You: Finding Likely Voters in Unlikely Places

Candidates come knocking on digital doors.

Using Big Data and voter data, political ad firms will find you. (Photo: SAUL LOEB/AFP/Getty Images)
Using Big Data and voter data, political ad firms will find you. (Photo: SAUL LOEB/AFP/Getty Images)

This year, my company will be responsible for tens of millions of dollars of online advertising, including the three biggest elections of 2013. I don’t care so much what website you’re visiting. I’ll find you, and all the other likely voters like you.

Sign Up For Our Daily Newsletter

By clicking submit, you agree to our <a rel="noreferrer" href="">terms of service</a> and acknowledge we may use your information to send you emails, product samples, and promotions on this website and other properties. You can opt out anytime.

See all of our newsletters

Using a few data points to target messages has been standard practice in politics for decades. Likely voters get robocalls, direct mail and visits at home from idealistic college kids looking to make a difference. If you’re not a potential voter, you’re dead to me. What’s new this time? Big Data and voter data got married and named their baby the National Online Voter File. The combination of the two spawned new voter-targeting ideas that give candidates literally millions of options.

The best part about all the news surrounding diffusion of President Barack Obama’s data machine is that it has created lots of demand by groups that don’t have the access to that brainpower. However, the news that the Obama Geek Squad generates is focused on its analytical prowess in minimizing inefficiencies with television advertising, not necessarily on its ability to deliver online advertising based on that data deep-dive. That’s where we come in: We take good data and use it to stick display and video ads on your laptops and mobile devices.

All you need to know is that audience targeting changed everything in politics. The concept is to advertise to an audience wherever they are, like likely voting, independent, college-educated women with kids in a specific zip code, versus buying ads on mommy blogs and hoping for the best. In practice, the Virginia gubernatorial election will be won or lost by who can best target ads to professional women in Arlington and Alexandria. No amount of ads in The Washington Post will make up for nonexistent targeting.

Here’s the math: Let’s say a gubernatorial candidate wanted to run online ads targeting Republican women who have voted in at least one of the last four elections. There are about 8.2 million people living in Virginia; however, only around a half a million of them are in that select group. Without targeting your message, 94 percent of the ads would be sent to the wrong people. That is where audience targeting comes in. We’re buying the audience of the half-million Republican women who have voted in one the last four elections.

That’s good news for the creative teams making great ads—they’ll be making an increasing number of them and delivering them to increasingly narrow slices of the electorate. The bad news is that making great political ads says nothing about your skill at making sure the right people see them.

The technological leap we’ve made is matching that voter file to the Internet so online ads can be as accurate as the phone calls, mail and knocks on your front door. Instead of buying ads on websites, we’re buying audiences. The technology is spreading, but you won’t notice the difference. While you’re seeing an ad for Christine Quinn, your unlikely-voting neighbor sees a Neiman Marcus ad. At the exact same time, an outside independent expenditure group that’s supporting Christine Quinn may have determined that your other neighbor, who is likely to vote but not for Ms. Quinn, might be seeing an ad for Bill de Blasio or against Bill Thompson, if the strategists and buyers have determined that a pro-Bill D or anti-Bill T vote helps Ms. Quinn’s goal.

What you will notice is that voter-targeted political advertising is now spreading through places you might not expect. Facebook opened up its walled garden to cookie-based advertising last year. We are voter-targeting inside LinkedIn and AT&T mobile devices.

Need to target likely-voting Democrats while they play “Words With Friends”? Reach low-income moms in target districts to weigh in on milk price-support legislation. What about likely voters who do not have health insurance? That’s now on the menu.

Most of the fancier mobile ads we’re handling are distributed through apps and games and mainly on the iPhone and Android. (Many players haven’t bothered with BlackBerry.) This is where the most innovation is occurring. Combining the National Online Voter File with geo-fencing lets us deliver ads that differ by neighborhood and voter type. When a Democrat enters Chinatown, he sees one ad; when an independent voter sets foot in Queens, she’ll get a different set of ads.  And when a likely voting Latina travels throughout the city, our ads follow her wherever she goes. Our partners at AT&T have been doing good work to create ad units for mobile devices and tablets that get your attention without making you want to pound your phone into the pavement.

BEING THE LARGEST voter-targeted online buying platform in the U.S. provides perspective. The change in attitude and understanding of online audience targeting happened virtually overnight. By August 2012, the majority of our online advertising programs were built around video ads rather than display ads. That month alone, 188 mil lion Americans watched online video, and, in some areas, publishers simply ran out of ads to sell. Think about it: The Internet literally ran out of space.

We did not handle the Mitt Romney online effort. That’s someone else’s fault. But what we did do was execute tens of millions of dollars worth of independent expenditures on Romney’s behalf for some of the largest Super PACs in the country. Inside that spend, we executed a multimillion-dollar buy in Ohio for one big player, which meant that another group that arrived hours later with hundreds of thousands of dollars meant for voter-targeted online video was out of luck. For a guy who earns a living selling ads at 1.4 cents each, that hurt.

Former South Carolina Governor Mark Sanford, just reelected to Congress, is the perfect example of why voter targeting matters. Very few people in his district weren’t sure of their opinion of him. We worked with Sanford’s former communications director, Joel Sawyer, who works for Push Digital to get his former boss through the 16-candidate primary, runoff and general election against Stephen Colbert’s sister. There are almost 700,000 residents of this congressional district; with help from their team, we narrowed it to the 30,000 likely voters who needed convincing that Mark Sanford was ready to complete his redemption story. This has been standard operating procedure for political mail for decades but new for online messaging.

What’s driving the accelerating shift in advertising toward video and mobile? Broadband now reaches 90 percent of homes with computers. More than half of mobile phones are smartphones. Less than half of households with incomes below $30,000 have broadband, so reaching your kids’ nannies on their MacBook Pro isn’t very likely, but that’s where mobile comes in.

Forrester Research predicts banner ads will nearly disappear within five years. In their place will be more sophisticated dynamic ads with functionality we haven’t thought of yet. Another evolutionary dead end in political technology is texting. It turns out that unsolicited text messages suck. And the only person in the universe with the charisma to drive millions of people to sign up for political texts is finishing his second term as president of the United States.

MOVING INTO THE campaigns this year, the vast majority of online political spending will end up on mundane sites where most of us spend our time. Now that I can find the same valuable likely swing voters on a sewing site, I’m not going to let any client spend real money on premium sites like Hulu or places that aren’t yet capable of voter targeting, like Pandora. The same philosophy that drove the Obama campaign to reach target voters on niche shows at 2 a.m. instead of during prime time we used to target online audiences wherever they decide to go.

The faster computers get, the more that can be done with the data exhaust from the many billions of ads we are delivering. This is also where the marriage of commercial and political data helps with traditional messaging; we can add color to what you think you know about your target group. Last month, we applied commercially available data to a list of supporters of a large advocacy group—assumed to be Fox News viewers—and found they spend their time watching CNN, golf and tennis and most definitely not Fox. New York State Democrats, we found, are five times more likely than the average American to eat Goya products, three times as likely to play racing video games, twice as likely to read someone else’s copy of The New York Times but half as likely to be a member of a union, church or synagogue. Combining political and commercial data allows us to target frequent drinkers to oppose a new tax on beer. And the same process allows us to specifically reach uninsured Americans and help drive down the acquisition costs for insurers.

As we enter the NYC mayoral race and New Jersey’s gubernatorial race, what more will you see this year and next? The velocity of voter-data updates is increasing—meaning, we’re able to stop running ads to voters once their early-voting ballot has been received. The data exhaust has more information than anyone knows what to do with. That’s where the innovation will come next: processing that data into something actionable, especially for clients who have less money than Obama (which is pretty much everyone). Our challenge, as we move into the next campaign cycle, is to manage the data integration that Obama’s reelection effort mastered. To be clear, it’s not a technological challenge; it’s dealing with egos and business interests that don’t always align with running the best campaign. If you see me on the train, there’s a good chance I’m on the way to NYC to walk some of our citywide clients through our targeting process; and if you see their ads, there’s a good chance you are one of their likely voters.

This is all new territory. The benchmarks are just now being set. Expected click-through rates, ideal video view lengths, expected donation rates, how to maximize ROI—all that stuff is nowhere near set in stone. Any shyster who says he knows is either a fool or liar. It’s as if we just discovered the world is round but have no idea what the other hemisphere looks like. The year 2013 is the new 1493.

Jordan Lieberman is president of CampaignGrid, which developed the first platform to deliver data-driven online advertising for political, advocacy and public affairs campaigns.  He is the former publisher of Campaigns & Elections magazine. 

Big Data Is Watching You: Finding Likely Voters in Unlikely Places