Dynamic Target User Selection Model for Market Promotion with Multiple Stakeholders

While recommendation platforms present merchants with a vast and transparent sales avenue, they have inadvertently favored dominant merchants, often sidelining small-sized businesses. Addressing this challenge, platforms are deploying multifaceted market promotion strategies both to help merchants identify potential users and to spotlight emerging items for users. A crucial aspect of these strategies is the efficient selection of target users. By channeling resources towards the most responsive users, there’s potential for a heightened return on marketing investments. In light of limited research in this domain, we put forth a tri-stakeholder considered user selection model with social networks (TriSUMS). This model recognizes the intertwined interests of three core stakeholders: merchants (items), platforms, and users. It harmonizes the objectives of these stakeholders through an integrated reward function and incorporates social networks to identify the prime target users for items of merchants adeptly. We validate TriSUMS using an exhaustive exposure user-item interaction dataset, assessed within a solid offline reinforcement learning framework.