Multi-level Revenue Sharing for Viral Marketing

In this paper we present the design and analysis of revenue sharing schemes for viral marketing over social networks. The increasing need for monetizing social networks more effectively is causing social network platforms to look for alternatives to online behavioral targeting. Specifically, we turn to cooperative game theory and the Shapley value to design revenue sharing schemes to incentivize users to help the social network platform for more effective viral marketing. Our goal is to identify mechanisms that achieve desirable objectives in terms of computability, individual rationality, and potential reach. In particular, we propose multi-level revenue sharing for referral-based and viral marketing over online social networks. We show via simulations that users have more incentive to collaborate with the social network platform in implementing the campaign when the revenue or discount is shared across multiple levels rather than the commonly used single-level model. For this purpose, we design the graph-based model, for which we show that computing the Shapley value is #P-hard. However, we show that in a variation of that model, which we call the tree-based model, computing the Shapley value becomes polynomial time. We also show that the revenue function is supermodular only in the tree-based model. Supermodularity of the revenue function entails desirable corollaries.