Research
What Is a Customer Worth to Algorithms?
Presentation: HBS Marketing Seminar 2024
Abstract: Firms use algorithmic decision-making for a spectrum of tasks, including recommender systems. However, while model development receives significant attention for improving performance, the importance of data collection often goes unnoticed. In marketing, customers serve a dual role: they are not only sources of revenue but also part of the training dataset for learning global preferences.To address this, we integrate interpretable machine learning methods to quantify the customers' information value in terms of improved algorithm performance. We analyze MovieLens data and construct a recommender system based on neural collaborative filtering. First, we find that by identifying the information value, firm can improve model performance by removing the low-value customers. Second, we find that customers' information value is not correlated with their activity volume but rather relies on the variation in outcome variables, such as the variance in ratings given by a customer.
Beauty in a Spectrum of Sizes: Sales Impact of Advertising Models’ Body Shape
with Chen Cao, Zijun Shi and Shunyuan Zhang
Presentation: Marketing Science Conference, Conference on Information Systems and Technology 2024
Abstract: In recent years, the emergence of more inclusive body sizes in fashion models has marked a transformative shift in the fashion industry, championing diversity and inclusivity. However, the consumer response remains ambiguous. This paper investigates the influence of model body size on product sales by analyzing a unique three-month longitudinal dataset from a large online marketplace in Asia. The data contains rich information on fashion products and consumer shopping behaviors such as pricing, transaction records, clickstreams, and images of advertising models wearing the product, enabling us to measure the models' body size and demographics. We find that a 1% increase in the advertising models BMI correlates with a 1.1% rise in sales, while such effect is less pronounced for relaxed-fit items. The sales lift is primarily driven by an increased likelihood of clicking product pages rather than increased conversion upon click. We also explore the underlying mechanism for our main findings. This research highlights that a tangible commitment to inclusivity can contribute to commercial success and hence provides significant implications for brands that wish to enhance inclusivity in designing or advertising their offerings.
Publications
Long-Term Care Insurance and the Well-Being of Older Adults and Their Families: Evidence from China
with Xiaoyan Lei, Chen Bai and Hong Liu, Social Science & Medicine, 2022, 296: 114745. [Link]
Presentation: International Health Economics Association pre-Congress Session, July 2021. [Slides]
Abstract: This paper examines the impact of China's public long-term care insurance (LTCI) pilots on the well-being of older adults and their families. Using panel data from the Chinese Longitudinal Healthy Longevity Survey, we exploit the rollout of the LTCI pilots across different cities during 2015–2017 and identify the effect of LTCI using a difference-in-differences approach. We find that the implementation of LTCI has reduced the likelihood of insured older adults reporting unmet activity of daily living (ADL)–related need for care, the intensity of informal care, and the amounts of ADL-related care expenditures and out-of-pocket medical expenditures. LTCI coverage was also associated with an improvement in self-reported health and a reduction in one-year mortality risk of the insured older adults. From a policy perspective, these findings suggest that the introduction of LTCI may benefit all members of the household, both care recipient and caregivers, which has important implications for the development of LTCI in China and other middle-income and developing countries.