Paper Number
ICIS2025-1850
Paper Type
Short
Abstract
This study investigates dynamic retargeting by comparing visual-based recommendation systems (VRS), which prioritize product visual similarity, with collaborative filtering recommendation systems (CF), which leverages user behavior and tends to increase diversity. Grounded in congruity and construal-level theories, we examine how alignment between customer preferences and retargeted content shapes responses across purchase stages. In a large-scale randomized field experiment, recommendation effectiveness is stage contingent. CF’s diverse suggestions perform better early, when customers are exploring, whereas VRS’ visually similar items outperform later, when preferences are narrower and concrete cues matter. Moreover, highly engaged customers respond more to VRS at later stages, and targeted promotions further amplify VRS, especially among cart abandoners. Together, the results clarify when and why VRS vs. CF works, and offer actionable guidance for tailoring retargeting by stage, engagement, and promotional context to maximize clicks and conversions.
Recommended Citation
Tsekouras, Dimitrios and Yang, Zherui, "Product Diversity and Visual Similarity in Retargeted Recommendations: A Field Experiment on Visual-Based Recommendation Systems" (2025). ICIS 2025 Proceedings. 18.
https://aisel.aisnet.org/icis2025/user_behav/user_behav/18
Product Diversity and Visual Similarity in Retargeted Recommendations: A Field Experiment on Visual-Based Recommendation Systems
This study investigates dynamic retargeting by comparing visual-based recommendation systems (VRS), which prioritize product visual similarity, with collaborative filtering recommendation systems (CF), which leverages user behavior and tends to increase diversity. Grounded in congruity and construal-level theories, we examine how alignment between customer preferences and retargeted content shapes responses across purchase stages. In a large-scale randomized field experiment, recommendation effectiveness is stage contingent. CF’s diverse suggestions perform better early, when customers are exploring, whereas VRS’ visually similar items outperform later, when preferences are narrower and concrete cues matter. Moreover, highly engaged customers respond more to VRS at later stages, and targeted promotions further amplify VRS, especially among cart abandoners. Together, the results clarify when and why VRS vs. CF works, and offer actionable guidance for tailoring retargeting by stage, engagement, and promotional context to maximize clicks and conversions.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
16-UserBehavior