Since Pat Caddell introduced message polling, the winning model for most down- ballot campaigns has been a one-size-fits- all message: Reduce your candidate to his/ her best attributes, the opponent to his/her worst op-research line items, and then hammer home the message—usually on TV. But this model came from a time when the media audience was easy to sort, categorize and find. Its greatest weakness? One-dimensionality.
For years, people have been saying the Internet is a game changer. But you haven’t seen it, have you? Sure it was great for Howard Dean and Barack Obama, but their methods haven’t exactly translated down the ballot, have they?
This narrative has tied the hands of new media types when we work with professionally run campaigns, because they tend to be salted with campaign pros who experimented unsuccessfully in the past. Until now, the campaigns willing to innovate have tended to be broke and hopeless. Democrat Al Franken was different. His Senate campaign in Minnesota hired an Internet director from the corporate world, Jon David Schlough, who didn’t buy into this narrative. Freed from the same old fight, we were able to focus on developing a new media way to leverage the most tried and true tool for down- ballots: paid media.
Before I get into the details, one last appeal to the non-believers. The modern Internet bears little resemblance to the Internet you experimented with in cycles past. Change is happening so quickly that half of what a four- year technical student learns their freshman year is outdated by their junior year. There were as many Google searches in the last two months of the 2008 cycle as there were in the entire two years of the 2006 cycle.
In 2008, both Harris and Pew reported that more than 80 percent of adults are regular Internet users, and according to an IDC report, the average Internet user spends nearly twice as much time online as they do watching television. Digital advertising will not replace television any time soon, but the huge chunk of us who are exposed to digital ads are the same people who have TiVo, so it’s time to take this inventory seriously. Some have. This cycle the savviest campaigns used search, context and display advertising for acquisition, by putting donation and e-mail sign-up appeals in front of people interested in their race. Unfortunately, this acquisition model is limited because it only grabs the “low- hanging fruit.”
Thanks to Democrat Al Franken’s Senate campaign, we now have a proven model to move beyond these strategies. We do it by tapping into the concept of the “long tail,” an Internet marketing theory popular in the corporate world. It’s based on the idea that the Internet audience is extremely fractured. So, instead of identifying the most universally persuasive messages and broadcasting them to a wide audience, in the long-tail model you take the most persuasive messages and nanotarget each one to the right niche.
People don’t go to one place, looking for one thing. Their whims take them to a million places. The trick is to be everywhere, with tightly targeted messages. It’s about showing them highly relevant factoids/ads tailored to the whim they’re currently indulging, which if clicked, will redirect them to a relevant part of your website or related off-site content. In short, long-tail nanotargeting takes those little gems—be it an endorsement, video, news story, or ask—and shows it to the people who would care. To this end, we ran more than 30 million impressions for the Franken campaign across five horizontal ad networks, two vertical networks and dozens of local news outlets.
We nanotargeted more than 125 niche groups, with more than 1,000 pieces of creative, for less than $100,000. On Google alone, an acquisition budget of less than $20,000 got us more than 20,000 clicks, 5,500 active e-mail sign- ups, and more than 2,500 donors. We were able to reach persuasion niches (this is akin to someone opening up and reading a mail piece) for a fraction of a penny per impression, and less than 50 cents per interaction.
What We Did—And How
It’s all about identifying niches. In addition to geographic and demographic targeting, on the Internet you can usually also target by keyword. This means you pay for certain words, and when people search for those terms or read something containing those terms, the ad is triggered. In real terms, Minnesotans who were searching for cheap gas or researching fuel-efficient cars saw ads about Franken’s plan to lower gas prices.
Second, it’s important to distinguish between persuasion and acquisition niches. With persuasion, you must first identify the most persuasive content available, then figure out who is most likely to be persuaded. When someone from your district searches for “SEIU Local 16” and you trigger an ad about the SEIU endorsement of your candidate, that’s a persuasion niche. With acquisition, you first identify the type of people you want to acquire, then figure out where they are and what to ask them for. When someone is reading an e-mail containing the phrase “ActBlue Newsletter” and you trigger an ad with a fundraising call to action, that’s acquisition.
Starting with persuasion: On the Franken campaign, we took every issue position, endorsement, the best videos and the best off-site articles, and built separate ad campaigns around each one. In addition to applying demographic and geographic filters, each campaign had its own keyword list (or lists), often containing hundreds of keywords. When we wanted to reach farmers, we started with keywords like “farm supply,” “feed stores” and “large animal veterinarian.” When we wanted to reach people with a military member in the family, we used words like “Duluth VA Hospitals” or “support our troops stickers.” When we wanted to reach people supporting an aging parent, we bought keywords related to eldercare.
Each week we acquired more data, which helped us find more keywords and refine the message. For example, the ad groups relating to gas prices were generating a ton of impressions (three times as many as Iraq), so we poured time into re%uFB01 ning both the message and the associated keyword list. We ended up with an ad that said “$1.40 Gas | Gas was $1.40 when Norm Coleman took of%uFB01 ce. “Read Al Franken’s plan here!” and a keyword list containing roughly 600 terms.
We started with text ads (which require no design), and tested as many different variations as we could while staying on message. Once we had statistically signi%uFB01cant data, we applied polling math and made more variations, rinsed and repeated. By the time we started actually de- signing ads, we had teased out exactly what was resonating. Think of the process as virtual focus grouping, using actual ad-returns.
We also ran TV spots as online pre-roll commercials. This was cool, because not only is video interruption advertising the most proven form of political advertising, but also we were able target more precisely online and use the returns to inform our buys.
Measuring success with persuasion is easy. Think of a click as the equivalent of someone opening and reading a direct mail piece. (The fact they clicked it is proof they “opened” the piece.) Assuming every mail piece you send is opened and skimmed, that means every well-targeted click you can get for less than the cost of one mailer is successful.
Impressions (the number of time your ad actually appears) are also important because of a psychological phenomenon known as source amnesia. Simply put, people remember “facts,” but not where they heard them. If someone reads a factoid in one of your text ads while skimming the headlines, that fact can become as real as the headlines themselves. (It’s the same phenomenon that makes attack ads work.)
With acquisition, we turned the process on its head. First, we identified who we wanted to acquire (such as donors, online activists, in-state activists, interns, etc.). Then we asked, “How can we find these people?” The “low-hanging fruit” are already searching for or reading information about the race. We were able to go farther by advertising to the long tail. With donors, the obvious target was people who give online to other Democrats, so we bought keywords like “Johnathan Zucker” (who signs Act Blue’s e-mails) and “David Plouffe” (Obama’s e-mails) in gmail adspace. Finding interns was a cinch, since students are all on Facebook. We targeted netroots activists through high-visibility %uFB02ash ads on liberal blogs. These were placed through Commonsense Media, a vertical (niche) ad network that services this community. As with persuasion advertising, we started with as many messages (asks) as we could think of, then used the data to tease out the most effective variations.
Measuring success with acquisition is even easier than with persuasion, because the major ad networks actually track conversions.
The biggest lesson learned from the Franken experience is that this type of advertising is a very slow burn. Keyword targeting relies on what people do right now (and farmers aren’t always looking for farm equipment). More- over, each week the ads are up means more data, which means more efficiency.
In real terms, we started our persuasion Google ads for the Franken campaign about two-and-a-half months out. We were able to run about 8.5 million nanotargeted impressions and drive about 10,000 direct voter interactions (either clicks or voluntary video views). This means most voters saw our ads three to five times and we were able to directly interact with 44 times as many voters as there were votes in the margin certified by the state, all for about 5 grand. But had we started with persuasion about 12 months out, our data suggests that through Google alone we could have shown more than 75 million impressions (more than 25 per Internet-using voter) and driven direct interactions with almost 3 percent of the electorate—for less than 50 grand.
Also, we learned there is real value in advertising off- site content. Simply put, ads for content on Time.com look like ads for Time.com—so the voter might not realize that they are being messaged to. Finally, if we had it to do over again, we would have spent more on Google and Facebook and much less on local newspaper placements. The former two represent the pinnacle of targeting technology. With Google, you can target extremely specific niches using geographic, demographic and keyword targets—and you can track the exact number of impressions, clicks and acquisitions. Facebook users report their age, location, gender and interests on their profile, which in turn can be used for ad targeting. The return on investment was staggering.
Newspapers were the opposite. Most demand expensive, %uFB02at-rate prices, and rarely offer much in the way of targeting. Moreover, newspaper websites have very fractured audiences. Readers may be there for a speci%uFB01c story, out of habit, or to the check the weather. It’s not an audience you can effectively micro-message to, unless your message is touting the endorsement of that specific paper.
One last word of caution: Before you go out and try this yourself, remember that advertising in a fractured audience means advertising across millions of websites. The top ad networks screen their publishers for objectionable content, but a few are bound to slip through the cracks. The risk can be minimized, but newcomers should tread carefully.
Josh Koster is the managing partner of Chong Designs LLC, a Washington, D.C.-based new media advertising firm that works with progressives.