MERIT: A Merchant Incentive Ranking Model for Hotel Search & Ranking
🤖 AI Summary
Current hotel search-and-ranking (S&R) systems on online travel platforms overly prioritize short-term platform revenue, neglecting incentives for merchant service quality. Method: We propose the first merchant-incentive ranking paradigm, addressing three core challenges—Matthew effect, ambiguous quality-performance mapping, and short-term vs. long-term objective conflict—via a monotonic Merchant Tower network that explicitly models the positive relationship between service quality and business performance, coupled with a multi-objective hierarchical pairwise loss function jointly optimizing platform revenue, merchant service quality (measured by Hotel Rating Score, HRS), and user experience. The model integrates service-quality features and HRS supervision signals in an end-to-end trainable framework. Contribution/Results: Offline evaluation significantly outperforms state-of-the-art baselines; online A/B testing shows a 3.02% improvement in HRS; the system has been fully deployed on Fliggy, serving millions of users and hundreds of thousands of hotel merchants.
Application Category
📝 Abstract
Online Travel Platforms (OTPs) have been working on improving their hotel Search & Ranking (S&R) systems that facilitate efficient matching between consumers and hotels. Existing OTPs focus on improving platform revenue. In this work, we take a first step in incorporating hotel merchants' objectives into the design of hotel S&R systems to achieve an incentive loop: the OTP tilts impressions and better-ranked positions to merchants with high service quality, and in return, the merchants provide better service to consumers. Three critical design challenges need to be resolved to achieve this incentive loop: Matthew Effect in the consumer feedback-loop, unclear relation between hotel service quality and performance, and conflicts between platform revenue and consumer experience. To address these challenges, we propose MERIT, a MERchant InceTive ranking model, which can simultaneously take the interests of merchants and consumers into account. We introduce information about the hotel service quality at the input-output level. At the input level, we incorporate factors of hotel service quality as features (as the underlying reasons for service quality), while at the output level, we introduce the metric Hotel Rating Score (HRS) as a label (as the evaluated outcome of service quality). Also, we design a monotonic structure for Merchant Tower to provide a clear relation between hotel quality and performance. Finally, we propose a Multi-objective Stratified Pairwise Loss, which can mitigate the conflicts between OTP's revenue and consumer experience. To demonstrate the effectiveness of MERIT, we compare our method with several state-of-the-art benchmarks. The offline experiment results indicate that MERIT outperforms these methods in optimizing the demands of consumers and merchants. Furthermore, we conduct an online A/B test and obtain an improvement of 3.02% for the HRS score. Based on these results, we have deployed MERIT online on Fliggy, one of the most popular OTPs in China, to serve tens of millions of consumers and hundreds of thousands of hotel merchants. To address these challenges, we propose MERIT, a MER chant I nceT ive ranking model, which can simultaneously take the interests of merchants and consumers into account. We introduce information about the hotel service quality at the input-output level. At the input level, we incorporate factors of hotel service quality as features (as the underlying reasons for service quality), while at the output level, we introduce the metric Hotel Rating Score (HRS) as a label (as the evaluated outcome of service quality). Also, we design a monotonic structure for Merchant Tower to provide a clear relation between hotel quality and performance. Finally, we propose a Multi-objective Stratified Pairwise Loss, which can mitigate the conflicts between OTP's revenue and consumer experience. To demonstrate the effectiveness of MERIT, we compare our method with several state-of-the-art benchmarks. The offline experiment results indicate that MERIT outperforms these methods in optimizing the demands of consumers and merchants. Furthermore, we conduct an online A/B test and obtain an improvement of 3.02% for the HRS score. Based on these results, we have deployed MERIT online on Fliggy, one of the most popular OTPs in China, to serve tens of millions of consumers and hundreds of thousands of hotel merchants.
Problem
Research questions and friction points this paper is trying to address.
Incorporating merchant objectives into hotel ranking systems
Addressing feedback-loop Matthew Effect and quality-performance relation
Balancing short-term and long-term revenue conflicts
Innovation
Methods, ideas, or system contributions that make the work stand out.
Defines Merchant Competitiveness Index for quality
Introduces Merchant Tower for ranking relation
Uses Multi-objective Loss for revenue balance
🔎 Similar Papers
No similar papers found.
