Christophe Van den Bulte is Associate Professor of Marketing at The Wharton School of the University of Pennsylvania. His research focuses on new product diffusion and social networks. He is the co-author, with Stefan Wuyts, of Social Networks in Marketing, published by MSI in 2007. Since then, Facebook has grown exponentially, as have other social media platforms. Earlier this year, Van den Bulte spoke with MSI about how network theory applies to these rapidly evolving challenges.
MSI: What are some of the main issues in network analysis that apply to word-of-mouth marketing?
Van den Bulte: The key idea of network analysis is that the pattern of connections matters. The issue that has attracted the most attention from marketers so far is that opinion leaders or influentials tend to have many connections in their networks. But there’s much more to network structure than that. For instance, do you have a well-integrated network, or are there several sub-communities with only a few bridging ties among them? In the latter case, those bridges can become very, very important to the spread of information, even if they are not “sociometric stars” with lots of connections.
Another important issue, and one that has not been leveraged much in marketing, is the “strength of weak ties.” New, non-redundant information is more likely to come from weak ties—people who are mere acquaintances. The problem is, those people have less of an incentive to help you out than your strong ties—people like friends and relatives who care about you. Job search is a classic example. People get surprisingly many new leads through their weak ties, but some of the most valuable leads come from strong ties. The reason is that sources want to share those attractive leads only with people they care a lot about. So, if we want to make “social search” on a network more effective, it would be good to have both kinds of ties.
A recent study of a German retail bank found that customers acquired through their referral program generated higher margins and lower churn rates. The difference in customer lifetime value implied a 60% ROI on the referral fee. Sears is currently experimenting with social coupons, predicated on exactly the same logic.
MSI: Are there some areas where network analysis could be applied, but the potential hasn’t been seen?
Van den Bulte: Definitely. Network analysis can also be applied within companies. This has generated very important insights and practical benefits in innovation management, an area where mobilizing knowledge and other resources spread throughout the firm is critical to success. Marketers, however, seem to ignore this, enamored as they are by customer networks. Sales force management and channel management are two other areas where network analysis can provide much insight, but have been under-utilized so far.
MSI: Do you think network theory offers a unique perspective on how you should measure social media?
Van den Bulte: A lot of the attention and dollars are moving to social networking sites like Facebook, but I don’t think that many people, including managers, really think about them as social networks. Many practitioners look at them just as glorified TV sets.
From a network perspective, you’d look at how influence operates along these connections or social ties, and you begin to think about reach beyond immediate ties. How quickly will something propagate into the network, or will it actually stop propagating at some point? There are some classic mathematical results in that area. How can I boost the virality of a campaign? If it goes from person A to B, what are the odds that it goes from B to C? What are the odds that it goes through a third connection? So, metrics reflecting this “global cascade potential” would be useful.
The notion of sub-communities matters here. Typically, social networks have some clumps of people with similar interests or physical locations. Are there bridges between these clumps or tribes? If there are, those bridges should be of particular interest to a marketer who wants to go from one community to another community. So, metrics reflecting the extent to which a network is integrated or balkanized, and metrics identifying those nodes that bridge the sub-communities may be under-utilized as of now.
One of the big problems nowadays in research using social media connections such as Facebook or Twitter is that we don’t know much about the salience and meaning of these ties. I’m connected to people on Facebook I’ve never met and I don’t care much about. How do you measure that? As more sub-communities and communities of interest emerge within Facebook or in other social media forums, we may be able to infer more about the meaning of those ties.
Van den Bulte: From an academic and practical point of view, contagion is a very big idea. The current key research question is, Why are people reacting to what other people do? Is it because of spreading awareness? Is it because it reduces their uncertainty? Is it because of normative pressure? Is it because of competitive concerns? Or loss of status? If we better understand those motives driving contagion, it will lead to better practice.
As I mentioned earlier, some easy and effective network marketing programs exploit homophily. As marketing academics catch up with practice, we will need to delve deeper into how homophily operates in commercial settings. For instance, why do people behave similarly if they’re connected to one another? Is this due to contagion? Or is it due to preference-based selectivity?
A classic example is smoking among teenagers. A teenager is more likely to smoke if his friends are smoking as well. One obvious explanation is that smoking is contagious: A teenager whose friends are smoking is more likely to start smoking. The alternative explanation is that people select their friends based on their smoking habits: A teenager who smokes is more likely to befriend another smoker.
This chicken-and-egg problem not only poses a technical challenge for research but also has substantive implications for practice. In one case, viral marketing works, in the other case, it does not. The odds are, of course that both contagion and selectivity are at work. So, contagion leads to similarity, and similarity leads to even more interaction and contagion. Fleshing out these dynamics will help us better understand how customers behave within their networks.