Product Promotion in Social Networks with Crowd Effect

Abstract

Product promotions on social media platforms are gradually replacing traditional advertisements. One major difference is the word-of-mouth effect, as people now have access to others' opinions through online platforms. We include this effect when we model the spread of information on social networks. Specifically, we consider how to select initial influencers and how to offer discounts to potential customers adaptively so that the total number of influenced people is maximized under two commonly used diffusion models, called the Independent Cascade model and the Linear Threshold model. Due to the additional global state variable, which, for example, can be used to capture the review summary at the time, we prove that both our problems are non-submodular. For the seed selection problem, we design an algorithm to iteratively tighten the estimation of the upper bound and get candidate seed sets. For the discount offering problem, we design a rolling horizon method with fluid approximation. We show the superiority of our methods against several benchmarks through numerical experiments. We also derive some managerial insights from our results. For example, we find that the marginal gain of adding discount budgets and total promotion time is generally decreasing. Besides, for short discount campaigns, we recommend that the platform should launch them in the early stages. Our methods can be directly implemented in general networks and help social media platforms make better decisions.

Keywords: Social Networks, Viral Marketing, World of Mouth, Algorithms