We’ve been working on developing model of influence and information flow in social media. I noticed on Data Mining a link to a very relevant and interesting paper paper on the dynamics of viral marketing.

The Dynamics of Viral Marketing, J. Leskovec (CMU), L. Adamic (Michigan), B. Huberman (HP).
“We present an analysis of a person-to-person recommendation network, consisting of 4 million people who made 16 million recommendations on half a million products. We observe the propagation of recommendations and the cascade sizes, which we explain by a simple stochastic model. … While on average recommendations are not very effective at inducing purchases and do not spread very far, we present a model that successfully identies communities, product and pricing categories for which viral marketing seems to be very e®ective.”

This image from the paper shows very different recommendation networks for a first aid study guide (left) and a Japanese graphic movel (right).

The paper’s key points, as summarized by Eric Kintz are

• Viral marketing does not spread well. In epidemics, high connectors are very critical nodes of the network and allow the virus to spread. In recommendations networks, a few very large cascades exist but most recommendation chains terminate after just a few steps.
• The probability of viral infection decreases with repeated interaction. Providing excessive incentives for customers to recommend actually weakens the credibility of those links. …
• Viral effectiveness varies depending on price and category. Social context has a high influence on the potency of viral infection. Technical or religious books for example had more successful recommendations than general interest topics. Smaller and more tightly knit groups tend to be more conducive to viral marketing.

The second point speaks to an issue that we’ve been trying to get a handle on — how and why influence varys over time.