Explicit Budget Optimization: Activating Traffic under Budget Constraints in Cold-Start Ad Bidding

Hongchang Wu, Size Wang, Weitong Ou, Zixin Shao, Feihong Liu, Yang Liu, Hongyan Xue, Nianhua Xie

Keywords: E-commercial bidding, Explicit Budget Optimization, Budget-constrained exploration, Traffic activation

Abstract: Online advertising platforms typically employ cost-constrained bidding frameworks that dynamically adjust real-time bids to optimize advertiser-defined objectives under strict cost constraints. These frameworks rely heavily on feedback signals, such as historical cost consumption and conversion outcomes, to regulate bidding behavior. While effective when sufficient feedback is available, they often struggle during the cold-start stage, where both conversion signals and reliable cost–value estimates are extremely sparse. As a result, bidding strategies tend to become overly conservative, leading to slow traffic activation and delayed signal accumulation, which further hinders subsequent optimization. To address this challenge, we propose Explicit Budget Optimization (EBO), a lightweight module that explicitly supports signal acquisition during cold-start through a small, dedicated exploration budget. Rather than optimizing immediate cost efficiency or advertiser value, EBO focuses on activating early-stage traffic and acquiring initial feedback signals under strictly bounded spending. EBO operates alongside existing cost-constrained bidding strategies and integrates with them via a dual-constraint bidding integration mechanism, in which the final bid submitted to the auction is selected as the maximum of the EBO bid and the base bid. This design enforces a lower-bound on traffic activation while preserving the original cost-control guarantees. Furthermore, EBO supports personalized exploration budgets across different advertisers, campaigns, and creatives, and employs a feedback-driven budget tracking and bid adjustment mechanism to ensure smooth and efficient consumption of the allocated exploration budget throughout the cold-start period. Extensive experiments conducted on Tencent’s advertising platform demonstrate that EBO significantly accelerates early-stage signal acquisition for cold-start campaigns while maintaining strict cost control.

Track: Long Paper

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 25