In early 2009, having just been hired to oversee the online component of Harry Reid’s re-election campaign, JD Schlough went to meet the Senate majority leader for the first time.
Reid knew how important the Internet had been to the Obama campaign in 2008. And, as a veteran of close races, he knew that he would need to maximize every strategic advantage to win what his team anticipated could be the toughest of them all.
“I want the biggest, best online campaign in the history of Nevada,” Reid told JD in that first meeting.
This is the story of how our firms built and deployed it.
A Commitment to Online
Reid’s commitment to online strategy was a necessary element in our success. Another was his fundraising machine, which made possible a generous campaign budget of $24 million. A third element, and the one we’ll focus on here, was the campaign’s trust in its digital strategists.
As soon as Well & Lighthouse was hired, JD sought out Josh Koster, and the Chong & Koster team got to work analyzing the playing field—both in Nevada and nationwide. Meanwhile, JD and Reid’s campaign manager, Brandon Hall, built their relationship. Per their agreement, JD would report directly to Brandon, taking full control of (and full responsibility for) online communications, fundraising, social networking, and digital paid media strategy. Together, they’d develop and deploy digital strategy alongside all the other elements of the campaign.
In addition to the usual functions of an online campaign, Brandon also approved a mission that other managers might find controversial: online voter persuasion. The online operation would look for cost-effective ways to take proven persuasion messages and add a digital component to the campaign’s strategy.
JD and Josh were convinced early on that producing “the biggest, best online campaign in the history of Nevada” meant finding new ways to get the Reid team’s message to voters—and Brandon’s trust in them gave them ample room in which to work.
Our strategy relied on three major components: Scaling long-tail nanotargeting campaigns across the entire electorate, measuring the buys not in clicks or impressions, but in points. Using the resulting paid traffic to build huge, highly segmented remarketing lists.
Taking advantage of emerging inventory in online television ads and flash banners to layer traditional media buys and show these proven messages to targeted audiences more often.
It’s one thing to find tiny audiences and niche-message to them. But even in a state with fewer than a million registered voters, finding ways to niche-message to the entire electorate across multiple networks and media—showing the right ad the right number of times to each person—is a challenge.
A slide from one of the animated ads the Reid team surged campuses with.
To understand how we did it, consider the online persuasion campaign we ran regarding healthcare reform.
When the controversial reform bill passed in March 2010, many Democrats shied away from promoting its benefits, preferring to change the subject to topics with broader appeal. Because Reid was so closely identified with the legislation, his campaign knew it didn’t have that option. Refusing to let the bill be defined by its opponents, they decided to own it; they fiercely defended reform, promoting its benefits to Nevadans, and generating a treasure trove of generally positive local coverage.
Owning it online was our job. If a Nevada political observer went online in the days and weeks following the bill’s passage, we would make sure they learned about the benefits of reform.
Not only that, we’d make sure they learned about benefits that were specifically relevant to them. This presented an enormous targeting challenge. To meet it, Tyler Davis at C & K started building models that would eventually become more complex than we ever anticipated.
In the weeks leading up to the vote, we had worked up eighteen different sets of creative—each containing Facebook ads, banner ads, and text ads—and nanotargeted each to a different segment of the electorate. We built a microsite with dynamic, parameter-based landing pages for each individual stream of traffic. Voters who saw and clicked on paid ads would land directly on a page describing the aspect of reform most relevant to them.
The day President Obama signed the bill into law, the Reid campaign went up with all eighteen different digital paid media tracks. As we launched creative over the next few days, we optimized each flight daily, using cheap Facebook ads to identify the best headline, image, and copy variation for each audience and turning what we learned into optimized banner and flash creative over the next twenty-four hours.
For nearly two weeks, we ran 400 points a day to each targeted voter. In other words, a twenty-five-year-old would see an ad about being able to stay on her parents’ insurance—four times a day, every day, for nearly two weeks. (Her parents would see that ad, too.) Seniors, on the other hand, saw ads about closing the Medicare donut hole. In short, anyone who clicked through landed on a page describing a benefit of the legislation specifically relevant to them. All in all, our healthcare reform effort ran 78 million impressions in Nevada. Over the course of the campaign, we did the same with issues such as foreclosure assistance, Yucca Mountain, veterans’ affairs, climate change—nearly a dozen in all. Each time we deployed one of these efforts, we got at least 50 million impressions worth of data, yielding nearly a quarter billion data points in all, which we mined and analyzed to create massive site lists for each voter subset.
As a result, we knew exactly who was responding to paid political ads online—and what kind of ads they liked. We knew which residual inventory contained voters—and exactly what we could afford to pay before it became inefficient. We knew which niche sites worked best and we knew exactly which sites within a given niche would perform before we launched a new campaign. We knew how to find all the same people on other ad networks, not just Google and Facebook.
Nanotargeting an entire electorate for persuasion might seem inefficient—but with diligent data collection and analysis, we taught ourselves to do it efficiently. By the end, we were launching nanotargeting “surges” in which the very first impression was more optimized than the last impression from the previous surge. In other words, we used all of our data about which sites had users who were responding to messages about the Senate race to inform the targeting for each subsequent buy. We were able to zero in on thousands of placements that were being read by voters as well as which messages to use on each and how much to pay for each site.
Once we had gotten people’s attention, we kept track of them with a method called remarketing. Here’s how it works: We install code onto various sites and landing pages that allows us to stamp visitors with a sort of virtual invisible-ink stamp. The stamp contains information—demographic and sometimes psychographic—about the traffic.
For example, if you were a Nevadan who clicked on one of our political ads, we inferred that you were following the race—and we tagged you as such. If you clicked on an ad related to healthcare reform, that went on your stamp, too. Because the invisible ink in the stamp comes not from specific websites you might subsequently visit, but from the ad networks themselves, we could then follow you around the Internet and show you our ads on any site that was part of a given network even if we hadn’t specifically targeted the site.
By the time voters really started paying attention a few months before the election, we had more than 3 percent of the electorate on our remarketing lists. That’s an entire margin of error’s worth of voters to whom we could show 1,000 points a day—that’s ten impressions—of our message for just a few pennies each.
A frame from one of four animated banners the Reid team developed to push Yucca Mountain as an issue.
It’s important to note that most of our persuasion campaigns, even the early ones, were executed in conjunction with traditional media buys. We worked side-by-side with the Reid media team and often layered our online flights on top of broadcast and cable television buys, mirroring their messages and giving them additional frequency by running the ads online. Other times, we layered on top of mail. And, occasionally, we let our online video stand alone. The Reid campaign developed one of the largest bodies of television creative in political history. The idea was to make sure that, when we had a message that we knew worked and creative that expressed it well, every single voter saw it often enough, no matter how they consumed media. For instance, take instream (or pre-roll) ads, the online equivalent of television commercials. These are fifteen- and thirty-second spots that stream before short clips and during the commercial breaks of television watched online. When our firms first started using them in Al Franken’s 2008 Senate campaign in Minnesota, they were a novelty, and there simply wasn’t enough inventory available to really make an impact.
But times have changed. Americans have become accustomed to watching video online—and to the ads that now run before and during their online content. According to a Pew study, one in three American adults consumes the kind of content that contains these ads—roughly the same penetration as cable. But online video is much cheaper, making it easy and affordable to produce multiple versions of a given clip.
That is our model: Find voters who care about a message. Keep track of them as they surf the Internet. Devise innovative ways to show them targeted creative over and over.
To reach college students, we Google-surged campuses and bought every single available impression for students in campus networks.
We told everyone within fifty miles of Yucca Mountain that Sharron Angle wanted to dump nuclear waste in their backyards.
We found women and told them that Angle didn’t want to make insurance companies cover preventative care like breast-cancer screening.
Designing Persuasion Creative
We designed our persuasion ads according to very strict criteria. Since we needed the ads to sink in even if people didn’t click on them or watch them in full, we designed each slide to stand alone and wrote non-linear narratives that made sense even if you didn’t see them start to finish. For example, if you were looking at a three-slide animated flash ad, you would be just as likely to see slide three, slide one, and then slide two, as you would to see slide one, slide two, and then slide three. We also made sure the final frame was a full static ad that told the entire story in as few words as possible, since sometimes people notice banner ads only after the animation has finished.
Plugging into the Broader Campaign
Meanwhile, the Reid campaign and Nevada Democratic Party field staff had their own incredible operation up and running—and we worked with them early on to organize Latinos, veterans, women, and students. Organizers got their own locally built social media constituencies and their own branded site properties. For instance, the Latino organizers had a “Latinos for Reid” Facebook page and a bilingual microsite that featured Latino issues and news. We achieved critical mass with each constituency through highly targeted Facebook ad campaigns. This connected organizers with local activists and helped the activists to connect directly with each other. We built connections among our base of fans, gave organizers exposure for their posts, and created a natural feedback loop that helped posts remain in Facebook feeds longer and yielded lots of highly targeted impressions of effective content at no cost.
A slide from an animated banner targeted to veterans.
These early efforts would prove invaluable when it came time for early voting. GOTV for early voting can be dangerous if you’re on the wrong side of the enthusiasm gap, as we were in 2010; target impressions poorly, and you can end up motivating precisely the wrong voters. So we decided to play it safe, activating early voting reminders only for groups that broke for Reid by more than two-to-one and only buying the impressions we were absolutely positive would hit their mark.
The eighteen-to-twenty-five and LGBT audiences clearly fell into this category and were easy to reach. Young people spend a ton of time online, and pretty much every ad network has solid age-related demographic targeting. (Facebook’s, in particular, is flawless.) The LGBT community is also easy to find on Facebook, which provides information on relationship status, as well as on LGBT sites.
We had to be more careful targeting liberals. Behavioral targeting is great, but it is so new that accuracy varies widely from one ad network to the next. Geotargeting works, too, but Nevada Democrats are most heavily concentrated right around the Las Vegas Strip, which made it difficult to target our desired audience without wasting impressions on tourists and people who worked in the area but lived elsewhere. So we stuck primarily to Facebook, but also bought out advertising on available liberal blogs through Commonsense Media and bid on a hand-picked list of left-leaning sites in the Google network.
The last group we targeted for early voting was Latinos, who make up over a quarter of Nevada’s population. We had planned to use online ads for persuasion, but our opponent,with her extreme anti-immigration position, took care of that, driving Latinos so strongly into Reid’s camp that we ended up simply targeting them as part of our early voting push.
We ran Spanish- and English-language creative through Google, Facebook, websites associated with Univision and Entravision, and Batanga, the largest Latino ad network. We bought so much inventory from these networks that we had some of them scrambling to find more. Latinos saw GOTV ads everywhere—and instead of just clicking through to the standard polling place locator, they also could call a special toll-free number listed in the ad to speak to a bilingual organizer and arrange a ride to the polls.
In all, we ran tens of millions of GOTV impressions to Latino voters, driving tens of thousands of them to take action. As a result, Latinos comprised a greater portion of the Nevada electorate than they had in 2008—up from 12 percent to 15 percent.
As Election Day approached, we deployed our final persuasion strategy. Editorial boards across the state had unloaded on Sharron Angle, and we built a long list of Reid endorsements, including many from boards that generally prefer Republicans—and a library of devastating quotes along the lines of the Reno Gazette-Journal’s pronouncement that “Replacing Sen. Reid with Angle would be a disaster for Nevada.”
A freeze frame from a flash ad targeted to female voters.
Our first job was to make sure that readers had seen their paper’s endorsement. We bought out online ad inventory (even buying the homepage takeover and wrapper of the Las Vegas Sun’s website the day before the election) and used Facebook ads and Google display ads to get extra frequency.
Meanwhile, we had flash ad templates canned and ready, allowing us to quickly produce thirty-second flash ads that tied together devastating third-party quotes about Angle. In the campaign’s final forty-eight hours, we bombarded likely voters statewide with these ads, using all the data we’d accumulated over nearly two years to guide our targeting.
Lessons for Others
Not every campaign will enjoy the Reid campaign’s advantages: a huge budget, a top-notch team of staff and consultants, and a dream opponent. But every campaign manager and every digital strategist can benefit from what we learned. The most important lesson was that digital strategy can be used for voter persuasion. But we also learned that we have to stop thinking small, in terms of clicks and impressions, and start thinking like media consultants—in terms of points. Real, needle-moving buys are measured in points, and as Americans’ content consumption habits change, online is going to look, feel, and act a lot more like television. On the Reid campaign, digital was treated like all the other elements of the media tool belt. Just like television, mail, phones, and the rest, our good ideas were funded, and the ones we couldn’t prove effective weren’t.
And just as we couldn’t have done what we did if we hadn’t had a seat at the table, we couldn’t have done it in isolation, either. Without the press team generating clips, we wouldn’t have had good content to showcase. Without the field team building a great early voter plan, our GOTV efforts would have had nothing to plug into. Our efforts paid off only because the entire Reid campaign saw them as part of an integrated strategy.
We believe that our approach is applicable to any campaign. These techniques can scale down—and they can scale down the ballot. (Indeed, we’ve put them to work for dozens of other clients, ranging from city council candidates to national awareness campaigns for major retail brands). All you need is the right campaign structure—and the sort of commitment to your online program that Harry Reid made the first time he met JD, twenty-one months before pulling off the upset of the cycle.
Jon-David Schlough is the founder of Well & Lighthouse LLC. Josh Koster is managing partner of Chong & Koster. Andy Barr is president of Well & Lighthouse. Tyler Davis is a partner at Chong & Koster.