Retargeted Versus Generic Product Recommendations: When is it Valuable to Present Retargeted Recommendations?

• 成果类型: Article
• 署名作者: Wan, Xiang (Shawn); Kumar, Anuj; Li, Xitong
• 署名单位: Santa Clara University; State University System of Florida; University of Florida; Hautes Etudes Commerciales (HEC) Paris
• 刊物名称: INFORMATION SYSTEMS RESEARCH
• ISSN/ISSBN: 1047-7047
• DOI: 10.1287/isre.2020.0560
• 发表日期: 2024
• 页码: 1403-1421
• 关键词: online; networks; systems; IMPACT; LINKS; MODEL

摘要

Although the effects of algorithmic product recommendations on product sales are understood, the differential effects of retargeted recommendations (recommended products a user has previously viewed) versus generic recommendations (recommended products a user has not previously viewed) are unclear. We conduct a field experiment to empirically examine the relative effect of retargeted versus generic recommendations on product sales at different stages of users' purchase funnel. The product recommendations can affect sales by influencing the number of product impressions and their conversion rates (purchase probability conditional on impression). We separately estimate the effect of retargeted and generic recommendations on product impressions and conversion rates. We find that (i) generic recommendations increase conversion rates only in the early purchase funnel stage, but retargeted recommendations do not affect conversion rates, and (ii) both recommendations result in a higher number of impressions of recommended products. Overall, retargeted (generic) recommendations result in higher recommended and total product sales in the late (early) purchase funnel stage. We also conducted a controlled experiment on Amazon MTurk to unveil that retargeting (showing previously viewed products to users) drives the effect of retargeted recommendations. Our counterfactual simulations show that the retailer can obtain up to three percent higher product sales by applying our findings to the existing recommendation systems. Our research has implications for online retailers and the design of algorithmic product recommendation systems.