Beyond Single Slot: Joint Optimization for Multi-Slot Guaranteed Display Advertising

2026-05-23 · Source: cs.LG updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, E-commerce & Digital Commerce · Depth: Expert, long

Summary

A novel joint optimization framework is proposed for multi-slot guaranteed display (GD) advertising, moving beyond single-slot assumptions to address challenges like slot-level redundancy, contract imbalance, and exposure concentration. This framework formulates ad allocation as an offline bipartite matching problem, incorporating a contract roulette mechanism for slot exclusivity and Page View (PV) constraints for impression control. It also features a scalable allocation optimization algorithm for large-scale deployment. Online A/B tests on the Meituan advertising platform demonstrated significant improvements, including a 28.99% increase in Average Revenue Per User (ARPU) under 70% traffic and enhanced contract stability via DID analysis. Further experiments showed ARPU increased by 28.17% and Fulfillment Rate improved by 2.12% compared to the previous production baseline.

Key takeaway

For MLOps Engineers or Ad Platform Architects tasked with optimizing guaranteed display advertising in multi-slot environments, your current single-slot allocation methods are likely underperforming. You should consider adopting a joint optimization framework that models allocation at the page-view level. This approach, incorporating Page View constraints and a contract roulette mechanism, can significantly improve merchant ROI, platform revenue, and contract fulfillment robustness. Online tests on Meituan showed a 28.99% ARPU increase.

Key insights

Jointly optimizing multi-slot guaranteed display advertising via bipartite matching and specific constraints significantly boosts platform revenue and merchant ROI.

Principles

• Multi-slot ad allocation requires coordinated, page-view-level optimization.
• Fine-grained slot control prevents over-exposure and promotes balance.
• Contract exclusivity reduces redundant impressions and improves user experience.

Method

The framework uses an offline bipartite matching formulation with Page View (PV) constraints for slot-level exposure control and a Contract Roulette-based selection module for probabilistic filtering and adaptive bidword control.

In practice

• Implement Page View constraints to cap per-slot impressions.
• Use a contract roulette mechanism for one-to-many assignment conflicts.
• Employ adaptive bidword control for dynamic retrieval across slots.

Topics

• Guaranteed Display Advertising
• Multi-slot Ad Allocation
• Bipartite Matching
• Constrained Optimization
• Ad Platform Monetization
• Meituan

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Milliseconds to Match: Criteo's AdTech AI & the Future of Commerce w/ Diarmuid Gill & Liva Ralaivola — Modern ad tech relies on millisecond-speed AI, deep learning, and foundation models for real-time, privacy-conscious recommendations.
• Contextual Bandits for Maximizing Stimulated Word-of-Mouth Rewards — A contextual multi-armed bandit framework optimizes stimulated word-of-mouth by learning individual spillover probabilities in social networks.
• Meet travelers in the moments that matter with Search Campaigns for Travel. — Google unifies travel ad management within standard Search campaigns for simplified workflows and enhanced performance.
• Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search — Balancing immediate conversions with long-term item growth is crucial for e-commerce platform sustainability.
• SmartFont: Dynamic Condition Allocation for Few-Shot Font Generation — Adaptive multi-level condition allocation is crucial for balancing global structure and fine-grained local style in few-shot font generation.
• My Unfiltered Thoughts On ChatGPT Ads — OpenAI's future revenue pressures may lead to more aggressive, integrated advertising strategies.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.LG updates on arXiv.org.