Do we listen to opinion leaders?

By Ben Mattison

In the summer of 2001, the toymaker Hasbro launched a new handheld videogame called P-O-X. In the game, warriors fought invisible “alien infectors” that had escaped from a laboratory. A radio transmitter allowed kids to “battle” each other from up to 30 feet away.

As part of the introduction of the game, Hasbro hired a marketing firm to create a viral word-of-mouth campaign. Marketers used surveys to find the coolest kids in Chicago and then, seeking to harness their influence, gave those kids samples of the game to hand out to their friends. The campaign, by all accounts, was a success: the cool kids loved P-O-X, and within a few weeks Hasbro had sold one million units.

P-O-X wasn’t, ultimately, a big national hit: after the September 11, 2001, terrorist attacks and the anthrax attacks that followed, parents and retailers weren’t interested in a game about infection. But the campaign has been frequently cited in studies of word-of-mouth marketing.

The P-O-X campaign is one of many in recent years based on the idea that all of us (not just 12-year-olds) turn to influential “opinion leaders” in deciding what restaurants to try, what books to read, and what products to buy. If marketers can successfully target their campaigns at this subsection of the population—the cool kids among us—the effect of their dollars could be multiplied dramatically.

Ed Keller has made a career of helping companies market to opinion leaders. He co-authored the 2003 book The Influentials while the CEO of the market research firm RoperASW and is now the CEO of the Keller Fay Group, which consults on word-of-mouth marketing. He pegs the number of opinion leaders, or “influencers,” at about 10% of the population. “These are the people who are friends, neighbors, colleagues at work who, if you’re seeking to get up to speed on something, they are the people you call,” he says. “They are everyday people who keep up with things that are new and have wide social networks”—which means they have more sources for learning about new things, and more people to talk to about them.

The idea that opinion leaders have a disproportionate influence on the rest of the population has its roots in the work of sociologist Paul Lazarsfeld, who, starting with the 1940 presidential election, did a series of studies originally designed to show how the mass media—then, primarily radio—affected voters’ political opinions. When Lazarsfeld and his colleagues asked voters what had convinced them to vote the way they did, many responded that they had been influenced, not by radio or newspapers, but by friends or acquaintances. Lazarsfeld eventually developed a theory of a two-step flow of information: information moved from the mass media to opinion leaders, and then from opinion leaders to the rest of the population.

In the 1960s, this idea was expanded by Everett Rogers and others, who put it in the context of theories about the diffusion of innovation through the population. According to Rogers, the adoption of a technological innovation begins slowly and then accelerates suddenly as it is embraced by the majority. In this formulation, opinion leaders play a crucial role: understanding the potential of a technology and conveying that information to their wider social networks. They are responsible for the inflection point, the moment when the curve turns sharply upward and the technology—the television, the personal computer, the iPod—takes off.

In the years since this work, opinion leaders have been the subject of many studies in marketing science and other social sciences; in the last decade, Malcolm Gladwell’s The Tipping Point helped spread the idea among the public at large. At the same time, as the fragmentation of media made traditional advertising less effective, marketers have turned more of their attention to “buzz,” “viral,” and other word-of-mouth marketing, including campaigns that target opinion leaders.

How, then, to identify opinion leaders? Researchers looking at a small network often use a “sociometric” method, in which they ask every member of the network whom he or she approaches for advice, and then map the responses. In larger networks, researchers and marketers rely on “self-reporting”: a survey, often based on an opinion leadership scale developed by Charles King and John Summers in 1970, which had seven questions about how often respondents talk to friends and neighbors about a category of products and whether they tend to give or receive advice and ideas.

The Keller Fay Group has such a questionnaire; to be identified as an opinion leader—what the company calls a “Conversation Catalyst”—a survey respondent must indicate that he or she has a wide social network, keeps up with things that are new, and is frequently sought out for opinions. Keller Fay combines that information with an ongoing survey, which it has run since 2006, in which it asks consumers to keep diaries of all of their conversations about brands.

According to the company’s data, Conversation Catalysts, who are roughly twice as likely as the general public to talk with friends and neighbors about brands, talk about new products before the general population does. For example, when Nintendo released its enormously popular Wii gaming system in 2006, Keller Fay’s surveys showed that the number of Conversation Catalysts mentioning the brand rose sharply several weeks before discussion began to increase in the general population.

Duncan Watts, a sociologist who holds posts at Columbia University and Yahoo!’s research arm, questions how big a role opinion leaders really play in the movement of ideas through the population. “Intuitively, when something big happens, we want to attribute it to some special cause,” he says. “And it’s convenient: if you can just find the influencers, they’ll do your work for you. But the point I’ve been making in my research is that it’s really an illusion.”

Watts created a stir with a paper, published in 2007, in which he and Peter Dodds questioned the basis of the “influentials hypothesis.” The heart of the paper was a computer simulation, in which the two created a series of virtual social networks with different properties and then tested how an idea moved through them. In some cases, they “seeded” an idea with simulated opinion leaders, who had greater power to influence their neighbors; in others, the idea originated with ordinary members of the group.

They found that while seeding the idea with opinion leaders increased its ultimate spread in some cases, the propagation of the idea was affected much more by the global conditions of the network, and by one condition in particular: how easy it is to influence members of the network.

“If you have a critical mass of easily influenced people,” Watts says, “and it’s ready to be ignited, an influential person is more likely to ignite it than an ordinary person, but only by a little bit. And if the critical mass doesn’t exist, there’s nothing that anyone can do.”

Dina Mayzlin, an associate professor of marketing at Yale SOM, recently published research, with Harvard’s David Godes, aimed at getting direct evidence of the effect word-of-mouth marketing can have on sales. They also wanted to see if, and under what circumstances, opinion leaders are in fact better than others at propelling sales.

In the study, groups of consumers in cities around the country were enlisted to talk up Rock Bottom Brewery, a national pub chain, in conversations with friends and acquaintances. Some of the consumers were already loyal customers of the restaurant; others weren’t familiar with it at all. Some in each group were identified as opinion leaders using the King and Summers questionnaire.

When Mayzlin and Godes compared diaries kept by the consumers to sales figures, they concluded that, somewhat counterintuitively, the word of mouth initiated by those who weren’t fans of the restaurant had a larger effect on sales. The explanation, they hypothesize, is that loyal customers were more likely to have already told their friends and acquaintances about Rock Bottom Brewery before the campaign began. But the less loyal customers were able to reach new potential customers.

The study also found that self-reported opinion leaders were more effective than others at getting their friends and acquaintances to try Rock Bottom Brewery—but only if they were loyal customers. For the less loyal customers, being categorized as an opinion leader had little or no effect on their ability to drive sales.

The “good news” for marketers, Mayzlin and Godes write, is that the study showed that they can create word of mouth (WOM) that leads to sales. “The ‘bad news’ is that although our results suggest that the firm should be using its less loyal customers to create WOM, it may not be able to rely on the popular opinion leadership scale to identify the most effective, less loyal disseminators of WOM.”

The larger conclusion, Mayzlin says, is that opinion leaders are better than the rest of us at propelling word of mouth about brands. But the task of making use of them for marketing campaigns is complicated by the fact that they are discriminating consumers. “Opinion leaders talk about things that they are genuinely excited about,” she says.

Wharton’s Christoph van den Bulte has also been doing research aimed at refining the understanding of opinion leadership and the circumstances in which it is useful for marketers.

He raises two key questions: “Can we identify opinion leaders and tell in advance who they will be for a particular product or a new technology or management practice? And are there products, technologies, and settings, more broadly, where social influence is more important than in others?”

It makes sense to Van den Bulte, for example, that consumers would turn to opinion leaders for advice when they are particularly uncertain or when the consequences of making the wrong decision are large: in buying an expensive piece of electronics, say, as opposed to a particular brand of potato chips

Van den Bulte, together with Wharton’s Raghuram Iyengar and USC’s Thomas Valente, recently conducted a study looking at the place of opinion leaders in the adoption by doctors of a new pharmaceutical drug. The drug (which was not identified) seemed especially suited to opinion leadership by these criteria: It treated a life-threatening viral infection for which other drugs already existed, so the stakes for doctors in switching were high. And because the disease was chronic, it would take years for doctors to learn whether the new drug was having a positive effect—a situation of high uncertainty.

Because communities of doctors are relatively small, the study also provided an opportunity to compare the two major methods of identifying opinion leaders: through sociometric data (network mapping) and self-reporting. Van den Bulte and his colleagues identified doctors who treated this infection in three cities. Using surveys and prescription data, they measured the adoption of the new drug among these doctors over 17 months after its release.

The study confirmed that adoption of the drug was accelerated by “contagion” among doctors—that is, not just because of advertising or other global effects. And it found that opinion leaders did contribute more to the spread of the adoption than ordinary members of the network.

But opinion leaders identified sociometrically—that is, those identified by their peers as sources of information—were more effective than self-reported opinion leaders in driving adoption. In addition, sociometric opinion leaders adopted the drug earlier than self-reported opinion leaders, and were more likely to be affected by the behavior of those around them.

“Clearly, these two metrics are not capturing the same thing,” Van den Bulte says. “Maybe what we have been doing in marketing—these self-reported leadership measures—do not really capture leadership. They capture self-confidence, and that’s something really different.”

If self-reporting isn’t an accurate way to find opinion leaders, how do marketers, who are dealing with populations too large to survey sociometrically, go about reaching them?

One solution, Van den Bulte says, is to find a proxy for opinion leadership in a particular product area. A diaper company looking to reach opinion leaders among mothers might focus on women with kids in school, because they have been shown to have larger social networks. “Just by focusing on women with children who are going to school, you will oversample the women who are well connected and undersample the women who are poorly connected.”

That may mean giving up on finding the cool kids, the segment of the population that drives public opinion in a wide range of arenas, Van den Bulte says. “But that doesn’t mean that we can’t find opinion leaders, or at least good proxies for opinion leaders, in each product category. I think that’s the next step for marketers.”