Off the coast of New Zealand last year, a kayaker was drifting along quietly when a seal burst from the water and slapped him in the face with a large octopus. As it happens, the trip was funded by GoPro as part of a product launch, and that sucker punch was caught on video by one of the company’s cameras. When GoPro posted the clip to Facebook, it exploded, yielding a publicity bonanza.
It’s a marketer’s dream: What could be better than having viewers voluntarily send your branded content to friends (and “friends”), saying, “You gotta see this!” Chasing that dream, companies today are moving more and more of their media spend to social channels. It’s a natural evolution — business goes where the buyers are, and nowadays that’s on the platforms.
But what kind of content works best on social media? Opinions abound; evidence, not so much. “There’s been very little real-world research on this,” says Harikesh S. Nair, a professor of marketing at Stanford Graduate School of Business. The problem is data: Facebook could do it, but each commercial user on Facebook sees only its own metrics; there hasn’t been a way to draw general insights.
So Nair, along with Dokyun Lee of Carnegie Mellon and Kartik Hosanagar of the Wharton School, hooked up with an analytics firm that gathers daily performance figures for some 800 business users of Facebook. By pooling the data for more than 100,000 posts — and using novel machine-learning techniques to characterize their content — the researchers published a paper that offers real answers that companies can use to optimize their social media strategies.
When firms first tiptoed into social media, Nair says, they brought to it a mindset from old media, where the goal was to maximize “reach,” which is simply the number of people exposed to a brand message. The corollary online, it seemed, was to maximize one’s follower count — and companies employed a variety of inducements like coupons and free swag, not to mention sponsored posts, to amass followers.
But eyeballs didn’t equal engagement. “Early audits showed that awareness wasn’t enough,” Nair says. “Followers weren’t interacting with the content in any way.” Which isn’t surprising: commercials were always an annoyance, the price of watching TV shows for free. Why would anyone “like” a corporation, comment on its self-serving ads, or share them with friends?
“The focus,” Nair says, “then became not just getting exposure, but figuring out what to put in the message so users will want to engage with it. This generated a whole new industry called ‘content marketing.’ The question became, what kind of content do I need to reach what kind of user, to generate what kind of engagement — and toward what goal?”
The goal might seem obvious, but there are two distinct streams of social media marketing, Nair says, with different objectives. One, known as “performance marketing,” aims to generate immediate sales, or “conversion.” The other is “brand building,” where the goal is to connect with consumers in a more personal way, in hopes of earning their long-term loyalty.
Interestingly, the data showed that most firms used one or the other exclusively — only a few did both at once. “I don’t know why that is,” Nair says, “but it could be an organizational thing — where the teams responsible for social media in different companies are more aligned with the sales or marketing departments.”
To characterize each post, the researchers first hired workers on Amazon Mechanical Turk to evaluate a subset of about 5,000 posts and label content based upon soft attributes like humor and emotion or upon hard information like price deals. Then they used that labeled set to train a computer, with natural language processing and machine-learning algorithms, to power through the rest.
When they finally combined the content attributes with the engagement figures — the actual likes, shares, and comments for each post — they found stark differences in performance: “When a company says, ‘Hey, here’s a coupon for 20% off,’ it gets very little engagement,” Nair says. “When it uses what we call ‘brand-personality’ content — essentially, when it talks to users in ways that simulate a human being — it gets lots of engagement.”
Mix It Up
On one level, of course, we know we’re being played — the bantering, irreverent personality on the other end is a construct, a brand profile, an avatar — but our brains can’t resist. Marketers hack our basic urge to connect with others who are like us, to sort ourselves into tribes. “There’s a lot of research showing that efforts at persuasion are more credible when they seem to come from an individual person, a friend,” Nair says, “instead of, say, a corporation in Cincinnati.”
However, performance marketing still has a place. Nair thinks users might be reluctant to share such posts because they don’t want to appear to be shilling for a company or seem unduly excited about saving money. “People are hyper-aware of the optics on social media,” he says. “They think about how their actions will look to others.” But that doesn’t mean they won’t click through for the coupon.
There’s also no reason a company can’t do both, he adds, and that may be where most of them fall short today: “A possible content strategy would be to combine them, to use informative posts to generate immediate leads and personality-driven posts to build long-term brand capital. A portfolio approach.”
Engagement Is King
Still, they’re not equally important; the study shows that the crucial element in any social marketing campaign is the brand-personality content. And the reason for that, Nair says, gets at the key difference between traditional broadcast media and new media: In this new world, user engagement is noticed by the algorithms that drive reach.
“Facebook’s News Feed algorithm gives priority to content with good engagement. If you’re sending out posts that aren’t getting any visible traction, the algorithm will say, ‘Aha, users don’t like this company’s posts very much.’ Then you get buried farther and farther down in News Feed, so you have little chance of ever showing up in front of users.”
In this way, Nair says, maximizing engagement — those precious likes, shares, and comments — on social apps is perhaps today’s equivalent to search engine optimization, or SEO, on the web. In both cases, the goal is to reverse-engineer the ranking criteria used by systems that essentially equate popularity with value.
“It sounds paradoxical, but you can’t get engagement without engagement,” Nair says. “So even if you’re mainly interested in performance marketing and driving sales, it’s really essential to add some good brand-personality content — if only to remain visible in this crowded marketplace.”