How to Identify Influence Leaders, Opinion Leaders in Social Media
How to Identify Influence Leaders in Social Media: Zsolt Katona
The currency of social media is influence. Credit-card companies offer rewards to customers with a high influence score, airlines give such people free flights, and some employers make job offers dependent on those ratings.
So how do they find opinion leaders? How do they determine a person’s influence? Where do these scores come from?
Recent research I conducted with Peter Zubcsek and Miklos Sarvary examined personal influences in an online social network in a European county. Building on an almost 300-year-old discipline, graph theory, we studied how members attracted their friends to join the network.
We found that, besides the number of one’s friends already using the service, the interconnectedness of these friends is just as relevant in making a person join. In other words, whether your friends know one another makes a big difference in how much they are able to influence you. This force is so strong that when five of your friends are already users, the effect of adding one friend as a user is the same as having one extra friendship between your friends.
Network structure matters. But how do we identify the influencers? Because we had access to data on all the friendships in the network, in addition to some demographic variables, we were able to determine how one person’s position in the network affected his or her power of influence. A common measure of influence is simply the number of connections someone has, but that isn’t always the correct way to pinpoint leaders.
We found that highly connected individuals have less influence on each of their friends on average. Having more friends can make your total reach and influence higher when it comes to compelling others to sign up for a free service, for example. But this isn’t necessarily so when it’s a matter of swaying people who are making an important decision. If you are in the process of buying a new car or a home, you are likely to ignore the advice of your highly connected but not so close friends. The lesson is that more connections don’t always mean higher influence.
To determine a person’s influence, it isn’t enough to simply count the number of connections they have. More meaningful are the communities or groups that people belong to within social networks. Most users have a separate but potentially overlapping set of friends that includes work colleagues, family and schoolmates.
To understand the mechanics of a social network it is imperative to be aware of these communities, but they are often impossible to observe. For example, in telecommunications networks, all calls are recorded, resulting in vast amounts of data on one-to-one connections, but no information on how the callers know each other.
We discovered that social-network analysis makes it possible to find these communities using network data, and we used our algorithm to identify communities in a number of phone networks.
Finding communities in a social network allows companies to identify influential individuals and to better understand their role. Some people are at the center of communities, possibly having a strong influence on their peers. Others may be on the periphery, but still have an important role in connecting different communities. Mapping out communities helps us understand how information travels in a network.
We conducted an experiment that illustrates the benefits of community-based social-network analysis. We obtained call data records from a mobile-phone provider in Asia that was trying to expand its loyalty program. Using our social-network analysis methods, we identified a target group of people who were active in a community for a text-message campaign that asked customers to join the program. We then compared the response rate to a control group that was targeted by the provider using only traditional marketing tools.
There was a substantial difference in response rates (9.2 percent versus 5.4 percent), proving the benefits of detailed network analysis. Moreover, a unique code sent with the messages allowed us to track customers who received a forwarded text message from a friend and not directly from the provider. Once again, our target outperformed the control group (6.3 percent versus 4.1 percent).
With all the network data available to marketers, it is important to appreciate the value in finding influential consumers and communities. We are just at the beginning of a long journey to fully understand the role of influence in social networks, but it’s already clear that simple metrics aren’t sufficient.
(Zsolt Katona is an assistant professor of marketing at the University of California, Berkeley’s Haas School of Business and a contributor to Business Class. The opinions expressed are his own.)