Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search

2026-05-18

AI Summary

This work addresses critical challenges in e-commerce search, including insufficient exposure of new items, misalignment between cold-start objectives and business metrics, and the lack of long-term growth evaluation. To tackle these issues, the authors propose GrowthGR, a novel framework that explicitly models and integrates item long-term growth value into the retrieval stage. Specifically, the ItemLTV module quantifies the incremental long-term transaction value of a single user interaction via counterfactual inference, while the MultiGR module incorporates multi-stage cascaded signals within a semantic ID generative retrieval architecture. Furthermore, a Multi-value-aware Policy Optimization (MoPO) strategy jointly optimizes short-term conversion and long-term growth objectives. Deployed on Taobao, the framework achieves a 5.3% increase in GMV for new items and a 0.3% overall search GMV uplift, significantly enhancing platform ecosystem health.

Abstract

New item growth is critical for maintaining a healthy ecosystem in large-scale e-commerce platforms. However, existing systems tend to prioritize presenting users with already popular items, a phenomenon often referred to as the "Matthew effect". In the context of search retrieval, current cold-start models suffer from the misalignment between training objectives and online business metrics, and they lack effective mechanisms to measure an item's growth potential. In this paper, we propose a Multi-Value-Aware retrieval framework tailored for e-commerce search, designed to better align with the cascaded online values across different stages of the search system while balancing immediate conversion and long-term item growth. Our framework GrowthGR consists of two key components: an Item Long-term Transaction Value Prediction (ItemLTV) module and a Multi-Value-Aware Generative Retrieval (MultiGR) module. First, in the ItemLTV module, we employ counterfactual inference to quantify the long-term value increment attributable to a single user interaction. Second, in the MultiGR module, building upon a semantic-ID-based generative retrieval architecture, we leverage structured samples with the search cascade signals and adopt a Multi-Value-Aware Policy Optimization (MoPO) training paradigm to align with multi-stage online values, while explicitly balancing short-term transactional value and long-term growth potential estimated by ItemLTV. We successfully deployed GrowthGR on Taobao's production platform, achieving a substantial 5.3% lift in new item GMV while delivering a non-trivial 0.3% gain in overall search GMV. Extensive online analysis and A/B testing demonstrate its positive impact on the overall ecosystem value.

Problem

• cold-start
• Matthew effect
• item growth
• e-commerce search
• long-term value

Innovation

• Multi-Value-Aware Retrieval
• Item Long-term Value Prediction
• Generative Retrieval
• Counterfactual Inference
• Policy Optimization