Decoupled Entity Representation Learning for Pinterest Ads Ranking

2025-09-04

AI Summary

To address the challenges of coupled user/Pin embeddings, delayed updates, and poor multi-task compatibility in Pinterest’s personalized recommendation and ad delivery systems, this paper proposes a decoupled upstream-downstream entity representation learning framework. The upstream model employs a deep neural network to asynchronously integrate heterogeneous behavioral and content signals, generating and periodically updating static user and Pin embeddings. Downstream tasks—including ad retrieval and CTR/CVR estimation—directly reuse these precomputed embeddings as input features, enabling modeling decoupling and cross-task representation sharing. This design balances representational expressiveness with system scalability. Extensive offline multi-task evaluations and online A/B tests demonstrate significant performance improvements across key metrics. The framework has been fully deployed in Pinterest’s production advertising system, driving substantial gains in core business metrics.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

In this paper, we introduce a novel framework following an upstream-downstream paradigm to construct user and item (Pin) embeddings from diverse data sources, which are essential for Pinterest to deliver personalized Pins and ads effectively. Our upstream models are trained on extensive data sources featuring varied signals, utilizing complex architectures to capture intricate relationships between users and Pins on Pinterest. To ensure scalability of the upstream models, entity embeddings are learned, and regularly refreshed, rather than real-time computation, allowing for asynchronous interaction between the upstream and downstream models. These embeddings are then integrated as input features in numerous downstream tasks, including ad retrieval and ranking models for CTR and CVR predictions. We demonstrate that our framework achieves notable performance improvements in both offline and online settings across various downstream tasks. This framework has been deployed in Pinterest's production ad ranking systems, resulting in significant gains in online metrics.

Problem

• Develops decoupled entity embeddings for user-Pin relationships
• Enables scalable asynchronous upstream-downstream model integration
• Improves ad ranking performance for CTR and CVR predictions

Innovation

• Decoupled upstream-downstream embedding framework
• Asynchronous entity embedding refresh for scalability
• Embedding integration in ad retrieval and ranking