AI in ads ranking - Multi-Stage Ranking of ads with increasing granularity

This article presents a novel approach to ad ranking in digital advertising platforms facing the challenge of evaluating hundreds of millions of potential ads for each user interaction. Current systems typically employ multi-stage ranking models that progressively filter ads based on increasingly complex criteria, optimizing for conversion or click probability. However, these approaches often fail to leverage the natural hierarchical structure of digital advertising—where ads belong to campaigns, which in turn belong to advertisers. This article proposes a hierarchical ranking framework that incorporates user affinity at all three levels, creating an aggregated weighted scoring system that considers ad-user interactions and campaign-user and advertiser-user relationships. This article promises more stable rankings, better captures contextual information embedded in campaigns, and acknowledges the distinct brand value different advertisers bring to similar products.