题　目：Dynamic Impact and Strategic Optimization of Sustainability Certification on E-Commerce Platforms

嘉　宾：Xiaohang (Flora) Feng (冯筱航), Ph.D. Candidate, Carnegie Mellon University

主持人：张铄 副教授 上海交通大学安泰经济与管理学院

时　间：2025年10月22日（周三）9:00-10:30

地　点：上海交通大学 徐汇校区 安泰经济与管理学院A303

内容简介:

Sustainability is becoming increasingly important in consumer behavior and platform strategy. Correspondingly, Amazon launched the Climate Pledge Friendly (CPF) program, where badge coverage, the share of green products receiving CPF, is a key policy lever. Too much coverage risks lower credibility; too little limits consumer choice and discourages seller participation. Empirically, we find that CPF raises demand and prices but lowers review volume, creating trade-offs in platform design and seller strategy. To study these dynamics, we estimate a dynamic structural model in which small-brand sellers jointly decide whether to adopt a sustainable strategy and whether to adjust prices, while anticipating future returns. By a sustainable strategy, we mean both obtaining an external sustainability certification and modifying product images and descriptions to highlight sustainability attributes. To simulate realistic counterfactuals, we must model how sellers adapt their content to highlight sustainability features. However, past structural models treat content as static, and off-the-shelf generative tools produce noisy, economically uninterpretable outputs. We bridge this gap by establishing a new Utility-Aware Contrastive Loss that incorporates economic goals to fine-tune CLIP (Contrastive Language–Image Pretraining), a model that learns joint representations of text and images. We then establish a utility-aware generative AI pipeline that transforms non-certified listings into badge-ready content that preserves semantic coherence, brand continuity, and sustainability information. Using six months of panel data on 1,200+ products, we evaluate what badge coverage best supports platform goals. For Amazon revenue, the optimal policy is stricter when more green products are present in a category, with stable badge visibility in the 10–30% range. For consumer welfare, the optimal badge coverage is about 20%. For seller welfare, full coverage is not optimal; the best outcome occurs at 80% coverage. Overall, the framework provides actionable insights for Amazon’s sustainability strategy, and the methodological approach generalizes to other multimedia platforms.

演讲人简介：

Xiaohang (Flora) Feng is a Ph.D. candidate in Marketing at Carnegie Mellon University, where she is profoundly grateful to be advised by Professor Kannan Srinivasan and deeply appreciative of the long-term mentorship of Professors Shunyuan Zhang and Xiao Liu. Her research integrates computer vision, generative AI, and machine learning to address marketing challenges, leveraging unstructured data, explainable AI, and human experimentation. She also studies sustainability and platform strategy using analytical modeling, causal inference, and structural econometrics. Her work has been published in the Journal of Marketing Research and the Journal of Consumer Research, with additional projects at an advanced stage at Marketing Science. Her research has been featured in outlets such as The Wall Street Journal and Harvard Business School Working Knowledge, and she has contributed book chapters on AI in marketing.

欢迎广大师生参加!