Bidding-Aware Retrieval for Multi-Stage Consistency in Online Advertising

2025-08-07

AI Summary

In online advertising cascade architectures, the retrieval stage lacks real-time bid signals, causing objective misalignment with the ranking stage and harming both platform revenue and advertiser performance. To address this, we propose Bidding-Aware Retrieval (BAR), the first framework to systematically incorporate bid information into retrieval. BAR ensures economic logic consistency via bid-aware representation learning; enables market responsiveness through an asynchronous nearline inference mechanism; and disentangles user interest from commercial value using task-specific attention and monotonicity constraints, further enhanced by multi-task knowledge distillation for joint optimization. Deployed at scale on Alibaba’s display advertising platform, BAR increases platform revenue by 4.32% and lifts positive operational ad impressions by 22.2%.

Technology Category

• Search and Optimization: Distributed Search
• Machine Learning: Online Learning & Bandits
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Ad search and search for Web retail
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems

Abstract

Online advertising systems typically use a cascaded architecture to manage massive requests and candidate volumes, where the ranking stages allocate traffic based on eCPM (predicted CTR $ imes$ Bid). With the increasing popularity of auto-bidding strategies, the inconsistency between the computationally sensitive retrieval stage and the ranking stages becomes more pronounced, as the former cannot access precise, real-time bids for the vast ad corpus. This discrepancy leads to sub-optimal platform revenue and advertiser outcomes. To tackle this problem, we propose Bidding-Aware Retrieval (BAR), a model-based retrieval framework that addresses multi-stage inconsistency by incorporating ad bid value into the retrieval scoring function. The core innovation is Bidding-Aware Modeling, incorporating bid signals through monotonicity-constrained learning and multi-task distillation to ensure economically coherent representations, while Asynchronous Near-Line Inference enables real-time updates to the embedding for market responsiveness. Furthermore, the Task-Attentive Refinement module selectively enhances feature interactions to disentangle user interest and commercial value signals. Extensive offline experiments and full-scale deployment across Alibaba's display advertising platform validated BAR's efficacy: 4.32% platform revenue increase with 22.2% impression lift for positively-operated advertisements.

Problem

• Inconsistency between retrieval and ranking stages in online advertising
• Sub-optimal revenue due to lack of real-time bid access
• Need for bid-aware retrieval to improve platform and advertiser outcomes

Innovation

• Bidding-Aware Modeling with bid signals
• Asynchronous Near-Line Inference for real-time updates
• Task-Attentive Refinement for feature disentanglement