Messages les plus consultés

jeudi 30 juin 2011

Seeding Strategies for Viral Marketing: Targeting Opinion Leader or not ?

Viral marketing

Do you remember "The Tipping Point", Malcolm Gladwell's best-seller ? I'ts about how trends work. Gladwell writes : "In a given process or system, some people matter more than others". In other words, people like opinion leaders are the "spark behind any successful trend".

So, marketers argue that targeted viral campaigning is more effective than good old mass marketing. They spend a billon dollars a year targeting precious influentials.  Are Klout or PeerIndex our future tyrants? It is clear that word of mouth (WOM) have strong influences on the success of viral marketing campaigns. But, is it useful to target key people, like opinion leaders, (ie. two-step flow model hypothesis) or not ? Does a random process works as well (network model structure)? More precisely, what is the optimal seeding strategy?

 Two-step flow model             vs          Network model of influence

Watts and Dodds (2007) tell us that people most easily influenced have the highest  impact on the information diffusion. They recommend  targeting a critical mass of influenceable people, rather than influential. Their conclusion is "Under most conditions that we consider, we find that large cascades of influence are driven not by influentials but by a critical mass of easily influenced individuals. Although our results do not exclude the possibility that influentials can be important, they suggest that the influentials hypothesis requires more careful specification and testing than it has received

A forthcoming article from Hinz and his colleagues (*) revisits this question. They compare four seeding strategies through one field experiment (a social platform like Facebook) and one real-life viral marketing campaign involving more than 200,000 customers of a mobile phone service provider. They use a sociometric method for identification of opinion leaders, that is respondents are asked to name the people they turn to for advice.

What is the best seeding strategy ?

First of all, their empirical results show that the best seeding strategies is targeting central opinion leader ("hub"). It can be up to eight times more successful than other seeding strategies.

Secondly, peripherical leaders  (« bridges ») are a powerful second best, because their influence is higher than random seeding.

What is the process behind this phenomenon ?

Why ?

Central opinion leaders constitute attractive seeding points because they are more likely to participate in viral marketing campaigns.

Moreover, these highly connected individuals also actively use their higher reach (ie. message diffusion on a higher number of person)


Central opinion leaders do not have more influence on their peers (ie. to make others to participate) than do less well-connected individuals.

Managerial implications

Marketers can improve the effectiveness of their viral campaigns by targeting central or peripherical opinion leaders (via sociometric method). Hinz et al. add : "Adding metrics related to social positions to customer relationship management databases is likely to improve targeting models substantially"..

And now our question is : how to identify an opinion leader in a social network, like Twitter or Facebook ? Klout or Peer Index metrics ? Self-designating method ? Sociometric method ? Results from Iyengar, Van den Bulte and Valente tend to promote sociometrics rather than self-assessment methods.

But no researcher has yet compared Klout or Peer index with traditional method...

(*) Oliver Hinz, Bernd Skiera, Christian Barrot & Jan U. Becker, Seeding Strategies for Viral Marketing: An Empirical Comparison, Forthcoming: Journal of Marketing, scheduled: January 2012.

3 commentaires:

  1. Ralph-Christian Ohr7 août 2011 à 17:23

    Thanks for the pointer - very interesting research on influence theory. I have to admit I'm not too familiar with influential behavior within networks.

    At first sight there seems to be no contradiction between outcomes on influence and the fact that highly complex systems (incl. human behavior) cannot be precisely predicted - though people tend to think they might be able to.

    What do you think?

  2. Thanks for your fruitful comments!

    The underlying reference is Watt's book "Everything is Obvious, Once You Know the Answer: How Common Sense Fails Us"(New York: Crown Business, 2011)and a discussion on August 4, 2011 by James Heskett on the Harvard Business site (

    According to James Heskett, Duncan Watts have a negative opinion about "works that purports to identify causes and effects in complex, unique situations involving such things as tipping points and many of the phenomena examined by the Freakonomists. In fact, nearly all writing about management and behavioral economics that seeks to credit performance to one cause or another is suspect. Anything based on this faux knowledge, including our common sense, is challenged."

    For instance, Watts & Dodds (2007 conclusion is that large cascades of influence are driven not by influentials (ie opinion leaders), but by ordinary folks. Influence in social network is mainly a random process. Hence,influentials hypothesis (ie Katz & Lazarfeld, Two step flow model) doesn't seem to work.

    Yes, but another research (Hinz & al, forthcoming in 2012) shows the opposite: Marketers can improve the effectiveness of their viral campaigns by targeting central or peripherical opinion leaders.

    Hence, contrary to Watt's opinion, we are able to predict (only in some case, I agree) the success of some marketing strategy in complex environment.

    Managers desperately need of academic research !

  3. Ralph-Christian Ohr7 août 2011 à 20:44

    Your research indicates that it's beneficial to leverage network opinion leaders for marketing strategies. It also shows that strategies should target at influentials, rather than e.g. at easily influenced people to become more effective / successful.

    But isn't this a relative statement, rather than an absolute and precise prediction of the strategy outcome?
    Example: if the wrong 'product' is launched (not validated until the launch has been finished), addressing opinion leaders might be most promising approach. However, the overall result will be poor anyway.