Brands spend billions of dollars a year on elaborate efforts to establish and maintain a social media presence. Think of the live-streamed video of a man setting a world record by skydiving from 128,000 feet (Red Bull) and the strange tweets sent from a supposedly hacked Twitter account that in fact originated with the company itself (Chipotle).
Facebook is the preferred platform: 80% of Fortune 500 companies have active Facebook pages. Each day enormous amounts of brand-generated content—articles, photos, videos, and so on—appear on those pages and on other social media platforms, all designed to entice people to follow, engage with, and buy from brands. Even the U.S. State Department seems enamored of acquiring followers, having spent $630,000 from 2011 to 2013 to garner Facebook likes.
Marketers often justify these investments by arguing that attracting social media followers and increasing their exposure to a brand will ultimately increase sales. According to this logic, recruits who socially endorse a brand by, for example, liking it on Facebook will spend more money than they otherwise would, and their endorsements will cause their friends (and friends of friends) to shop—creating a cascade of new business. At first glance the evidence seems to support this rationale: Many brands have discovered that customers who interact with them on social media do spend more money than other customers. A recent influential study by comScore and Facebook found that compared with the general population, people who liked Starbucks’s Facebook page or who had a Facebook friend who liked the page spent 8% more and transacted 11% more frequently over the course of a month.
Fatal logical flaw
Merely liking a brand on Facebook doesn’t change behavior or increase purchasing.
But that study and others like it contain a fatal logical flaw: They confuse cause and consequence. It’s possible that getting people to follow a brand on social media makes them buy more. But it’s also possible that those who already have positive feelings toward a brand are more likely to follow it in the first place, and that’s why they spend more than nonfollowers. In 23 experiments conducted over the past four years and involving more than 18,000 people, we used an A/B testing method to explore a crucial counterfactual: what followers would have done had they not followed a brand. Given the millions of dollars in marketing budgets that flow to social media at many companies, the distinction is not trivial. It has enormous implications for marketers’ resource allocations and for how they manage their brands’ social media presence.
In our experiments, we gradually added complexity to test four increasingly interactive ways in which Facebook might affect customers’ behavior. First, we tested whether liking a brand—that is, passively following it—makes people more likely to purchase it. Second, we examined whether people’s likes affect their friends’ purchasing. Third, we examined whether liking affects things other than purchasing—for example, whether it can persuade people to engage in healthful behaviors. Finally, we tested whether boosting likes by paying Facebook to display branded content in followers’ news feeds increases the chances of meaningful behavior change. We chose to use Facebook in our experiments because it is the dominant social network, but we believe that our findings apply to other popular platforms as well.
The results were clear: Social media doesn’t work the way many marketers think it does. The mere act of endorsing a brand does not affect a customer’s behavior or lead to increased purchasing, nor does it spur purchasing by friends. Supporting endorsements with branded content, however, can have significant results. And given that social media pages are gathering places for loyal customers, they can offer brands a unique source of customer intelligence and feedback from a crucial cohort. Armed with this knowledge, marketers can build new, more successful social media strategies.
Testing the Effects of Likes
Basic psychological principles give reason to suspect that liking a Facebook page could indeed change behavior and increase sales. Research has shown that people experience “cognitive dissonance” when their actions don’t reflect their beliefs, so it would stand to reason that a social media user who endorses a brand on Facebook would be more likely to buy it. Yet that’s not what we found.
In one of our first studies, conducted by two of us (Leslie John and Oliver Emrich) and our Harvard Business School colleagues Michael Norton and Sunil Gupta, half the participants were invited to like a new cosmetics brand on Facebook; most accepted. The other half did not receive this invitation. All participants were then given coupons for a free sample—redemption would serve as a proxy for purchasing. Members of the two groups were equally likely to redeem the coupon; it didn’t matter whether they had been invited to like the Facebook page or not. This finding held in subsequent studies, in which we increased the length of time between proffering the invitation to like and extending the coupon offer; it also held when we ran the experiment with a variety of new and existing brands. Across 16 studies, we found no evidence that following a brand on social media changes people’s purchasing behavior.
Measuring the Return on Facebook Like
In our second set of experiments, we sought to determine whether liking a page influences the behavior of online friends. When people like a brand on Facebook, their endorsement is typically broadcast to a subset of their network. Any subsequent engagement with the brand—likes, posts, comments, and shares—also appears in some of their friends’ news feeds. In classic marketing, word-of-mouth endorsements by peers have been shown to increase sales. But the value of endorsements may be lower on social media, for a couple of reasons. First, on many platforms, including Facebook, Twitter, and Instagram, following does not guarantee brand exposure for either endorsers or their friends. Facebook’s algorithms determine what content appears in a user’s news feed, and a user’s liking of a brand is broadcast to only a very few friends (without this intervention, users would be exposed to an average of 1,500 posts daily). Second, some Facebook users appear to like brands indiscriminately or for various one-off reasons—to get a discount, say.
To test the effects of social media endorsements, we asked 728 people who had recently liked a brand for the e-mail addresses of three friends. We sent each friend a coupon for one of the brand’s products, varying the information provided about the referral. In each group, one person was told that his or her friend liked the brand in the conventional, offline sense and had sent the coupon. The second person was told that his or her friend liked the brand on Facebook and had sent the coupon. The third person was told only that his or her friend had sent the coupon; people in this category made up the control group.