The Power of Linear Programming in Sponsored Listings Ranking: Evidence from Field Experiments

Haihao Lu, Luyang Zhang and Yuting Zhu

Papers from arXiv.org

Abstract: Sponsored listing is one of the major revenue sources for many prominent online marketplaces, such as Amazon, Walmart, and Alibaba. When consumers visit a marketplace's webpage for a specific item, in addition to that item, the marketplace might also display a ranked listing of sponsored items from various third-party sellers. These sellers are charged an advertisement fee if a user purchases any of the sponsored items from this listing. Determining how to rank these sponsored items for each incoming visit is a crucial challenge for online marketplaces, a problem known as sponsored listings ranking (SLR). The major difficulty of SLR lies in balancing the trade-off between maximizing the overall revenue and recommending high-quality and relevant ranked listings. While a more relevant ranking may result in more purchases and consumer engagement, the marketplace also needs to take account of the potential revenue when making ranking decisions. Due to the latency requirement and historical reasons, many online marketplaces use score-based ranking algorithms for SLR optimization. Alternatively, recent research also discusses obtaining the ranking by solving linear programming (LP). In this paper, we collaborate with a leading online global marketplace and conduct a series of field experiments to compare the performance of the score-based ranking algorithms and the LP-based algorithms. The field experiment lasted for $19$ days, which included $329.3$ million visits in total. We observed that the LP-based approach improved all major metrics by $1.80\%$ of revenue, $1.55\%$ of purchase, and $1.39\%$ of the gross merchandise value (GMV), compared to an extremely-tuned score-based algorithm that was previously used in production by the marketplace.

Date: 2024-03, Revised 2025-02
New Economics Papers: this item is included in nep-exp, nep-inv and nep-pay
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://arxiv.org/pdf/2403.14862 Latest version (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:arx:papers:2403.14862

Access Statistics for this paper

More papers in Papers from arXiv.org
Bibliographic data for series maintained by arXiv administrators ().