How Digital Inexperience Paid Off in the Trump Campaign

The answer for digital campaigners moving forward isn’t to abandon experiments, testing, and analytics. The answer is to listen better and to continuously test standing assumptions.

The 2016 Trump campaign’s digital director, Brad Parscale, was featured on 60 Minutes last week. Much of the interview focused on the central role of Facebook in Trump’s digital strategy. Parscale shared that he “understood early that Facebook was how Donald Trump was going to win. Twitter is how he talked to the people. Facebook was going to be how he won. …We did [ads] on Twitter, Google search, other platforms. Facebook was the 500-pound gorilla, 80 percent of the budget kind of thing.” He also revealed that the Trump campaign had been closely advised by Facebook staffers who were literally “embedded” within their offices. While this little fact led the news, the truth is that top tech platforms have been offering such services to political campaigns for years. What was news, however, was the revelation that the Clinton campaign had turned Facebook down.

Set aside all the Russia intrigue for a minute. There’s a basic puzzle here. The Clinton digital team was supposed to be one of the smartest, most sophisticated operations in history. The Trump digital team was, by their own admission, “just a few guys and a big Twitter account.” Trump invested $60 – 70 million in Facebook advertising, and used it as an online fundraising engine that brought in $240 million in online donations. How the hell did team Trump leapfrog team Clinton in the use of social media for campaigning?

The answer, I think, dates back to a digital experiment from the 2014 election, and the broader trend toward experimentally-informed campaigns. Facebook was offering some powerful new targeting tools in the 2016 race, including one you may have heard a little about, called “lookalike” audiences. Lookalike advertising sounds like a promise of magic beans, and the two campaigns had very different reactions to this promise.

Fundraising through Facebook advertisements was a sucker’s bet

Within the world of digital campaigning, there’s been a long-running debate between email specialists and social media consultants. For roughly a decade, social media marketers have announced every election cycle that “email is dead; social media is king.” Every cycle, professional email writers have replied, basically, “my email list against your social media advertising… wanna make a bet?” And every cycle, the social media marketers have mumbled an excuse and skipped out on the bet.

Facebook, Instagram, Pinterest, and Twitter have long been the showhorses of campaign communications. They’re meant to look nice and garner media attention. But ugly old email is the workhorse. If you want to raise money, you’re going to have to build a big email list, and then you’re going to have to bug that list for donations, and then you’re going to have to list-build some more, to replace the people who just unsubscribed from your list.

And this approach to fundraising via social media isn’t just campaign folk wisdom. As recently as the 2014 election, campaigns have run multiple experiments to determine the best and worst places to invest their digital dollars. Trilogy Interactive ran one such study, working with a Senate campaign to track the impact of Facebook advertisements and campaign emails on fundraising. Trilogy found that a $15,000 Facebook ad buy produced only $6,000 in direct Facebook fundraising (a net loss of $9,000). But their study also concluded that the Facebook ads produced $26,000 in additional indirect fundraising. In essence, the campaign’s email list was being primed by Facebook ads, leading supporters to become more likely to donate via email. Social media fundraising wasn’t a total waste of money for political campaigns, so long as it was looped into a multichannel communications plan still centered around email.

I heard about these findings at Rootscamp 2014, the annual post-election gathering that progressive organizers have been holding every cycle to share and learn best practices. Trilogy presented its experimental findings to a packed room at the Washington Convention Center. It produced a white paper that circulated among Democratic political operatives. The study confirmed that email was still the star of electoral campaigning, but Facebook could be part of a working ensemble cast.

That was the reality for digital campaigners circa 2014. Every cycle, social media vendors would make bold claims and promises in a bid to increase their share of a campaign’s communications budget. Every cycle was billed as the “YouTube election” or the “Facebook election” or the “Twitter election.” (Hell, 2016 was briefly billed as the “Meerkat election.”) Savvy campaigns sorted through these claims by running experiments, and smart campaign professionals paid attention to these experiments to inform their planning for the next election cycle.

Magic beans

One of the benefits of experimentally-informed campaigning is it can help campaigners sort through the bullshit. National elections in the United States are a multi-billion dollar industry, and that attracts a lot of sales pitches. In today’s polarized, gerrymandered political landscape, it is rarely possible to tell which specific communications tactics helped or hurt a candidate’s campaign. Every winning campaign has a thousand consultants explaining how they provided the critical insight that won. Every losing campaign has a thousand consultants explaining how things would have been different if they’d just gotten more funding and a nicer seat at the table. (I wrote about this trend for Civicist back in October 2016, in a post titled “Preparing for Campaign Technology Bullshit Season.”) Digital vendors and consultants can sometimes get away with offering magic beans to a new campaign, promising that the power of the internet has changed everything.

Two Yale political scientists, Alan Gerber and Don Green, launched the experimental revolution in political campaigning with their 1998 canvassing experiment and subsequent book Get Out the Vote!: How to Increase Voter Turnout. Their work spurred the formation of the Analyst Institute in Democratic politics, which in turn invited ongoing Republican attempts to establish their own data and experimentation shop. The whole point of experimentally-informed campaigns is to sort through all the noise and false consultant/vendor promises and establish what actually works.

Back before Trump was being treated as a serious candidate, the 2016 election was supposed to be the one when Republicans finally started to catch up with Democrats in their use of social science experiments in elections. Then Trump happened, and everything got, well, weird. Many Republican digital campaign professionals were active #NeverTrumpers, further isolating the Trump digital team from any established base of digital campaign knowledge.

As a result, the experienced digital politics professionals weren’t in the room for Trump when Facebook arrived with its marketing pitch. The Clinton digital team had seen the experimental results. They had been around for past cycles, and had heard all these bold social media promises before. Facebook was touting its new-and-improved lookalike advertising product and asking for a giant slice of the digital advertising pie. The data from past elections said otherwise. Parscale, meanwhile, effectively responded by saying, “Magic beans?!? Take all of my money!”

Google and Twitter sent embedded staff to the Trump campaign as well. And the Clinton campaign accepted some staff embeds from big tech firms, even if Facebook was not among them. The digital technology firms didn’t just have a seat at the table with Trump’s digital program; they were often the most knowledgeable and experienced voices in the room. That’s generally a terrible way to run a campaign. You’ll get sold a bill of goods more often than not.

Except this time, the beans turned out to actually be magic.

The half-life of digital campaign knowledge

If the 2016 campaign had been run using 2014’s Facebook advertising product, then the Trump campaign would have been setting its money on fire by spending 80 percent of its digital budget on Facebook ads. But in the intervening years, Facebook built a better ad platform. Campaigns can now target supporters and identify “lookalike” targets more effectively today than they could three years ago. Lookalike advertising is a powerful acquisition tool that helps a campaign identify and advertise to Facebook users who have similar demographic information or interests as a specified target population (fans of your Facebook Page or other custom audiences). Where campaign email professionals have to pursue costly acquisition strategies to acquire the email addresses of potential new supporters, Facebook’s lookalike advertising automatically extends the campaign’s message. So if you have a Facebook Page called “Gun Owners for Trump,” lookalike targeting will let you place ads both in front of fans of the page and in front of people whose Facebook behavior is similar to fans of the page. That’s radically different from Trilogy’s 2014 experiment, in which a Senate campaign was contacting its existing supporters through Facebook and email.

By 2016, Facebook had also improved its native analytics and A/B testing tools. When Brad Parscale brags about “testing 50 – 60,000 ads on an average day,” what he actually means is that they set up the Facebook ad platform to test thousands of color and language variants on their advertisements. A/B testing is well established in email campaigning, but until recently it was hard to replicate on social media.

Someday, I hope we will hear more details from the Clinton digital team about why they turned down Facebook. 2016 was a long campaign, and there were plenty of opportunities to try out the new advertising tools, observe results, and incorporate them into the digital strategy.

In 2016, the Trump digital team benefited from its lack of expertise. Expertise within the Clinton digital team led to a knowing mistrust of vendors. The lack of expertise in the Trump digital team led them to install vendors in key roles and believe their claims about the fundraising power of social media.

The answer for digital campaigners moving forward isn’t to abandon experiments, testing, and analytics. The answer is to listen better and to continuously test standing assumptions.

2016 won’t be the last time that we see meaningful changes in our digital environment. Knowledge of past experimental findings is an asset, so long as a campaign doesn’t treat this received knowledge as though it is set in stone. The answer for experimentally-informed campaigns isn’t to reject all future experiments, but to build more and better testing practices into our strategic processes.

(And if you’ve read this far, you should also consider reading my book, Analytic Activism, which discusses how a culture of testing and analytics can help social justice campaigners adapt and thrive in the digital age.)