HiBid: A Cross-Channel Constrained Bidding System with Budget Allocation by Hierarchical Offline Deep Reinforcement Learning

TL;DR

HiBid addresses cross-channel constrained bidding with dynamic budget allocation by framing the problem as a hierarchical CMDP and solving it with offline DRL. It introduces three key contributions: an auxiliary batch loss to balance channel budgets, lambda-generalization to adapt bids to budget changes, and CPC-guided action selection to satisfy cross-channel CPC constraints. Extensive offline and online experiments on Meituan data, plus a synthetic simulator, show HiBid outperforms six baselines in clicks, CSR, and ROI, and its deployment demonstrates practical effectiveness for large-scale advertising. The approach enables robust, budget-aware bidding across channels, improving advertiser outcomes while preserving platform revenue.

Abstract

Online display advertising platforms service numerous advertisers by providing real-time bidding (RTB) for the scale of billions of ad requests every day. The bidding strategy handles ad requests cross multiple channels to maximize the number of clicks under the set financial constraints, i.e., total budget and cost-per-click (CPC), etc. Different from existing works mainly focusing on single channel bidding, we explicitly consider cross-channel constrained bidding with budget allocation. Specifically, we propose a hierarchical offline deep reinforcement learning (DRL) framework called ``HiBid'', consisted of a high-level planner equipped with auxiliary loss for non-competitive budget allocation, and a data augmentation enhanced low-level executor for adaptive bidding strategy in response to allocated budgets. Additionally, a CPC-guided action selection mechanism is introduced to satisfy the cross-channel CPC constraint. Through extensive experiments on both the large-scale log data and online A/B testing, we confirm that HiBid outperforms six baselines in terms of the number of clicks, CPC satisfactory ratio, and return-on-investment (ROI). We also deploy HiBid on Meituan advertising platform to already service tens of thousands of advertisers every day.

Figures (9)

• Figure 1: Overview of the considered cross-channel constrained bidding (c$^3$-bidding) scenario.
• Figure 2: Hierarchical CMDP modeling for c$^3$-bidding.
• Figure 3: The proposed framework: HiBid
• Figure 4: Hyper-Parameters Tunning. The Capacity satisfactory ratio is calculated as the average of $\mathds{1}_{\left|\sum_{m=1}^M Cost(a^h_{m,p}) - \kappa_p \right| \leq \epsilon}$ of all channels and Budget satisfactory ratio is calculated by the average of $\mathds{1}_{\sum_{i=1}^I Cost(a^l_{p,i})\leq a^h_{p}}$ of all advertisers, respectively.
• Figure 5: Offline performance comparison with 5 baselines in terms of $Click$ and $CSR$ improvements.