Content Promotion for Online Content Platforms with the Diffusion Effect

Abstract

Problem Definition: Content promotion policies are playing an increasingly important role in improving content consumption and user engagement for online content platforms. However, a frequently used promotion policy generally neglects employing the diffusion effect within a crowd of users. In this paper, we study the candidate generation and promotion optimization (CGP) problem for online content through incorporating the diffusion effect. We also investigate ways to use the content adoption data to estimate the diffusion effect and to optimize the content promotion decision.
Methodology: We propose a diffusion model for online content that captures platform promotions. Using the diffusion model to characterize the diffusion process, the CGP problem is formulated as a mixed-integer program, which turns out to be NP-hard and highly nonlinear. Furthermore, taking advantage of the rich data available from the online platforms, we propose double ordinary least squares (D-OLS) estimators for diffusion coefficients.
Results: We prove the submodularity of the objective function for the CGP problem, which yields an efficient $(1-1/e)$-approximation greedy solution. We show that the D-OLS estimators are consistent and more efficient (i.e., with smaller asymptotic variances) than the traditional OLS estimators. Numerical results based on real data from a large-scale video sharing platform show that our diffusion model effectively characterizes the adoption process of online content. In addition, our proposed content promotion policy can increase the total adoption by 22.48\% when benchmarked against the policy actually implemented on the platform.
Managerial Implications:
We not only highlight the essential differences between the diffusion of online content and that of physical products but also provide actionable insights for online content platforms to substantially improve the effectiveness of a content promotion policy by leveraging our diffusion model.

Keywords: Online Content, Diffusion Modelling, Promotion Optimization, Approximation Algorithms