Robust autobidding for noisy conversion prediction models

2025-10-09

AI Summary

In online advertising auctions, prediction noise in CTR/CVR estimation induces bias in automated bidding strategies, jeopardizing advertisers’ ROI and platform robustness. To address this, we propose RobustBid—the first bidding framework that incorporates closed-form robust optimization into conversion-rate–driven bidding. RobustBid explicitly models uncertainty in CTR/CVR predictions and proactively defends against adversarial perturbations. It employs a lightweight, analytically tractable optimization scheme, balancing theoretical rigor with scalability for large-scale production deployment. Extensive experiments on the iPinYou dataset, real-world BAT (Baidu, Alibaba, Tencent) traffic data, and synthetic benchmarks demonstrate that under high-noise conditions, RobustBid significantly increases total conversions (+12.3% to +18.7%) while reducing average cost-per-click (−9.5% to −14.1%), outperforming both non-robust baselines and state-of-the-art risk-aware methods such as RiskBid. The approach exhibits strong practicality, computational efficiency, and system-level scalability.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

Abstract

Managing millions of digital auctions is an essential task for modern advertising auction systems. The main approach to managing digital auctions is an autobidding approach, which depends on the Click-Through Rate and Conversion Rate values. While these quantities are estimated with ML models, their prediction uncertainty directly impacts advertisers' revenue and bidding strategies. To address this issue, we propose RobustBid, an efficient method for robust autobidding taking into account uncertainty in CTR and CVR predictions. Our approach leverages advanced, robust optimization techniques to prevent large errors in bids if the estimates of CTR/CVR are perturbed. We derive the analytical solution of the stated robust optimization problem, which leads to the runtime efficiency of the RobustBid method. The synthetic, iPinYou, and BAT benchmarks are used in our experimental evaluation of RobustBid. We compare our method with the non-robust baseline and the RiskBid algorithm in terms of total conversion volume (TCV) and average cost-per-click ($CPC_{avg}$) performance metrics. The experiments demonstrate that RobustBid provides bids that yield larger TCV and smaller $CPC_{avg}$ than competitors in the case of large perturbations in CTR/CVR predictions.

Problem

• Addresses uncertainty in CTR and CVR prediction models
• Proposes robust autobidding to prevent large bid errors
• Improves conversion volume while reducing advertising costs

Innovation

• RobustBid method for autobidding with uncertainty
• Leverages robust optimization against CTR/CVR perturbations
• Derives analytical solution for runtime efficiency