Reorder Prediction and Advertising Strategy Design for Retail Media Network Product Ranking

Reorder prediction has been widely studied as a classification problem in e-commerce, but little research has explored how to use the predicted reorder probabilities for downstream product ranking. Extend reordering prediction to Retail Media Network (RMN) advertising in this paper and consider sponsored products as competitors to organic recommendations for display slots. A LightGBM model trained on user-level, product-level, and user-product interaction features achieved a mean AUC of 0.6771 in five-fold cross-validation and 0.6968 on a held-out test set, and consistently outperformed logistic regression. Personalised Top-10 recommendation lists were generated for the organic baseline. A simulated RMN environment compared the four advertising strategies in terms of product selection, bid generation, and advertisement quantity. Each strategy was evaluated according to the quality of recommendations (Precision@10, Recall@10, NDCG@10) and advertising results (exposure rate, average ad rank, simulated revenue). As a result, the strategy that preserved the recommendation quality better (NDCG@10 = 0.8904) had a lower revenue (109.19), whereas the revenue-maximising strategy (542.40) resulted in more disruption (NDCG@10 = 0.7108). It was not due to the selection of products or the quantity of advertisements; random bidding unexpectedly had a lower average ad rank (2.9956) than conversion-based bidding (around 4.97). Users with weaker organic recommendation signals were more likely to be shown the ads. Based on the above results, an empirical foundation for the design of RMN strategy has been provided; platform operators should choose a position on the recommendation-revenue trade-off curve explicitly, and per-user monitoring needs to be introduced to prevent excessive harm to vulnerable users.

Authors

Institutions

Publication Details

Journal
Advances in Economics Management and Political Sciences
Published
2026-11-17
DOI
https://doi.org/10.54254/2754-1169/2026.bh39170
Primary Topic
Recommender Systems and Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Reorder Prediction and Advertising Strategy Design for Retail Media Network Product Ranking

Yuxuan Dong
1 links
Advances in Economics Management and Political Sciences
Recommender Systems and Techniques
article

Reorder Prediction and Advertising Strategy Design for Retail Media Network Product Ranking

Yuxuan Dong
article en

Abstract

Reorder prediction has been widely studied as a classification problem in e-commerce, but little research has explored how to use the predicted reorder probabilities for downstream product ranking. Extend reordering prediction to Retail Media Network (RMN) advertising in this paper and consider sponsored products as competitors to organic recommendations for display slots. A LightGBM model trained on user-level, product-level, and user-product interaction features achieved a mean AUC of 0.6771 in five-fold cross-validation and 0.6968 on a held-out test set, and consistently outperformed logistic regression. Personalised Top-10 recommendation lists were generated for the organic baseline. A simulated RMN environment compared the four advertising strategies in terms of product selection, bid generation, and advertisement quantity. Each strategy was evaluated according to the quality of recommendations (Precision@10, Recall@10, NDCG@10) and advertising results (exposure rate, average ad rank, simulated revenue). As a result, the strategy that preserved the recommendation quality better (NDCG@10 = 0.8904) had a lower revenue (109.19), whereas the revenue-maximising strategy (542.40) resulted in more disruption (NDCG@10 = 0.7108). It was not due to the selection of products or the quantity of advertisements; random bidding unexpectedly had a lower average ad rank (2.9956) than conversion-based bidding (around 4.97). Users with weaker organic recommendation signals were more likely to be shown the ads. Based on the above results, an empirical foundation for the design of RMN strategy has been provided; platform operators should choose a position on the recommendation-revenue trade-off curve explicitly, and per-user monitoring needs to be introduced to prevent excessive harm to vulnerable users.

Advances in Economics Management and Political SciencesVol. 311(1)
Shaanxi Normal University (CN)
Recommender Systems and Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon