Michelle (Yingze) Song

Michelle (Yingze) Song
PhD Student, Marketing
PhD Program Office Graduate School of Business Stanford University 655 Knight Way Stanford, CA 94305

Michelle (Yingze) Song

I am a PhD candidate in quantitative marketing at Stanford Graduate School of Business. My expected graduation date is June 2022. I use causal inferences and structural methods to examine various important questions in quantitative marketing. Currently, my research focuses on recommendations, advertising auctions, consumer inertia, and digital marketing. Prior to joining Stanford, I completed my undergraduate studies at University of Virginia.

Research Interests

  • Recommendations
  • Advertising Auctions
  • Digital Marketing
  • Consumer Inertia

Job Market Paper


Personalized recommendations are known for their ability to navigate shoppers to the most relevant products first, saving their time. However, the hidden cost is that shoppers are less likely to find other desirable products along the search process serendipitously. Such a potential cost casts doubts on whether websites should adopt personalized recommendations. I suggest a positive spillover effect of gained efficiency from personalized recommendations: consumers explore more because increased search efficiency countervails an increasing opportunity cost of time. In addition, the total shopping time is likely to decrease if efficiency gains prevail over enhanced exploration expectations. I examine these hypotheses empirically using field experiment data from Missfresh, one of China’s biggest grocery delivery platforms. Personalized recommendations enable consumers to reduce the search for essential items, spend more time exploring other categories, and make more purchases while decreasing their total shopping time. These findings are important because they show consumers' active exploration under time pressure and they demonstrate a demand-increasing mechanism of increasing search efficiency through personalized recommendations.

Working Papers


Bidding in search advertising is commonplace today. However, determining a bid can be challenging in light of the complexity of the auction process. By designing the mechanism and aggregating the information of many bidders, the advertiser platform can assist less sophisticated advertisers. We analyze data from a platform that initiated a bid recommendation system and find that some advertisers may simply adopt the platform's suggestion instead of constructing their own bids. We discover that these less sophisticated advertisers were lower-rated and uncertain about ad effectiveness before the platform began offering information through the recommended bids. We characterize an equilibrium model of bidding in the Generalized Second Price (GSP) auction and show that following the platform's bid suggestion is theoretically sub-optimal. We then identify sophisticated and less sophisticated advertisers' private values using observed bids and the disclosed information. Counterfactual results suggest that the ad platform can increase revenue and the total surplus when it shares more information. Furthermore, the hybrid of auto-bidding with manual bidding could be a more efficient mechanism as we substitute less sophisticated bidding behavior for algorithmic bidding. These results shed light on the importance of exploring interactions between sophisticated and less sophisticated players when designing a market.

Work in Progress

Consumer Brand Inertia, Network Effects, and Persistence of Brand Market Share in Emerging Digital Markets

with Wesley Hartmann