Brij Disa Centre for Data Science and Artificial Intelligence
We investigate the informativeness of hygiene signals in online reviews, and their effect on consumer choice and restaurant hygiene. We first extract signals of hygiene from Yelp. We find that reviews are more informative about hygiene dimensions that consumers directly experience -- food temperature and pests -- than other dimensions. Next, we find causal evidence that consumer demand is sensitive to these hygiene signals. We also find suggestive evidence that restaurants that are more exposed to Yelp are cleaner along dimensions for which online reviews are more informative. Our results have implications for government inspections when consumers rate providers online.
About the Speaker
Dr. Hema Yoganarasimhan is a Professor of Marketing at the Foster School of Business, University of Washington. She also holds affiliate appointments in Computer Science and Engineering, Department of Economics, and Center for Statistics in the Social Sciences. Hema serves as a co-editor at Quantitative Marketing and Economics and as an Associate Editor at Marketing Science and Management Science.
She is recognized as one of the leading experts in quantitative marketing. Hema’s research brings together large-scale marketing data, economic theory, and econometric and machine learning tools to help firms optimize and automate their marketing decisions.
Her recent work focuses on combining machine learning tools and statistical econometric methods to address important problems in the domain of digital marketing and online platforms. One stream of research focuses on targeting in mobile and online advertising – how to target ads at scale using personalized user history and how can we quantify the optimal level of targeting from a platform’s perspective? In another ongoing project, she develops methods to personalize search rankings in online platforms in real time. She has also done work on estimating the role of reputation on sellers’ and buyers’ behavior in online auctions. Together, her recent body of work presents creative yet technically viable solutions to the challenges that businesses face in today’s world.
Hema’s research has won many prestigious awards, including the MSI Alden G. Clayton Doctoral Dissertation Proposal Award, Frank M. Bass Outstanding Dissertation Award, and John D.C. Little Best Paper Award. She has also been recognized as a “MSI Young Scholar” in 2015, a "MSI Scholar" in 2020, and an Erin Anderson Emerging Female Marketing Scholar and Mentor in 2021.
Hema received her Ph.D., M.A., and M.Phil. in Marketing and Business from Yale School of Management. Prior to that, she received her bachelor’s degree from Indian Institute of Technology (IIT) Madras. She has been at University of Washington since 2014, where she teaches MBA and undergraduate classes on analytics for the 4Ps of marketing and an advanced PhD class on dynamic structural models.
Online marketplaces use ranking algorithms to determine the rank-ordering of items sold on their websites. The standard practice is to determine the optimal algorithm using A/B tests. We present a theoretical framework to characterize the Total Average Treatment Effect (TATE) of a ranking algorithm in an A/B test and show that naive TATE estimates can be biased due to interference. We propose a bias-correction approach that can recover the TATE of a ranking algorithm based on past A/B tests, even if those tests suffer from a combination of interference issues. Our solution leverages data across multiple experiments and identifies observations in partial equilibrium in each experiment, i.e., items close to their positions under the true counterfactual equilibrium of interest. We apply our framework to data from a travel website and present comprehensive evidence for interference bias in this setting. Next, we use our solution concept to build a customized deep learning model to predict the true TATE of the main algorithm of interest in our data. Counterfactual estimates from our model show that naive TATE estimates of clicks and bookings can be biased by as much as 15% and 28%, respectively.