An Interview with Mengxia Zhang
How would you describe overarching concepts behind this article in the most simplified way?
In this paper, we investigate whether consumer-posted photos can serve as a leading indicator of restaurant survival above and beyond reviews and other known factors. We analyzed 755,758 photos and 1,121,069 reviews posted on Yelp between 2004 and 2015 for 17,719 U.S. restaurants with machine learning techniques. We find that photos can help forecast restaurant survival above and beyond reviews and other known factors. Interestingly, the informativeness of photos seems to play a bigger role than the aesthetics of photos.
What was the motivation to speak specifically to this matter?
This research idea was conceived in my second year in the Ph.D. program.
In my spare time, I like to take photos of food and restaurants and share them with my friends. Oftentimes, before I go to a new restaurant, I also like to check the photos posted by other consumers on Yelp to figure out whether I like that restaurant and what dishes I want to order.
This made me wonder: Is there any relation between consumer posted photos and restaurant survival? If so, what aspects of photos matter the most? May it be the content of photos, aesthetic quality, or other aspects? Further, If content matters, what type of content would play a bigger role? Would it be the photos depicting food, interior, outside, or other aspects of a restaurant?
To systematically answer these questions, I started to work on the project “Can Consumer-Posted Photos Serve as a Leading Indicator of Restaurant Survival?”. In particular, I use data from Yelp.
Who are your collaborators?
Lan Luo. Lan Luo is an Associate Professor of Marketing at the Marshall School of Business, University of Southern California. She was my advisor during the Ph.D. program.
Where are your collaborators located?
Los Angeles
Is this part of an ongoing collaboration?
My research focuses on artificial intelligence (AI) and new media. This paper is also about how to use AI to better understand new media like consumer-posted photos. Lan and I also have working papers on consumer-AI co-creation and the double-edged sword role of social networks.
Do you have future plans for this research? If so, what are they?
I plan to continue doing research about AI and new media. One on hand, topic-wise, natural extensions of photos include videos and virtual reality. I have working projects related to these topics. I am also interested in how consumers interact with AI. On the other hand, methodologically, I think machine learning has a great potential for marketing research. I am interested in exploring utilizing machine learning methods for different marketing questions.
We are grateful that Yelp made a large amount of its data public for research. We also thank Nvidia (a company that provides GPUs for running deep learning models) and Clarifai (a computer vision API for understanding photos) for supporting this research.
Did you have any particular influences with this specific content/article
Three factors motivate our research.
- Consumer-posted photos are ubiquitous, but we have yet to distill managerial relevant information from them. With the extensive use of camera-enabled smartphones and the increasing popularity of various photo-sharing platforms, three billion photos are shared on social media daily (McGrath 2017). The number of photos taken by consumers in 2017 was projected to be 1.3 trillion globally (New York Times, July 29, 2015).
- The restaurant industry is well-known for its high turnover rate, and it is not clear whether consumer photos can help predict restaurant survival. According to Parsa et al. (2005), the first-year turnover rate of restaurants is as high as 26%. Nevertheless, large-scale empirical research on restaurant survival has been scarce. Historically, it has been well documented in the business survival literature (e.g., Lafontaine et al. 2018; Parsa et al. 2005) that firm characteristics, competition, and macro conditions are the main factors associated with business survival. Many businesses with consumer-posted photos also receive consumer reviews that contain rich information about consumers’ descriptions and/or opinions toward their consumption experiences. As such, it is not evident that consumer-posted photos would play a role in predicting business survival after all these alternative factors are controlled for. On the flip side, consumer-posted photos may contain unique information that cannot be directly captured by firm characteristics, competitive landscape, macro environment, or consumer reviews.
- Multiple stakeholders can readily apply our proposed framework to improve their decision-making processes. Business investors and landlords can employ our research results to obtain a better evaluation of the market as well as to monitor restaurant survival likelihood. Our models can significantly increase survival prediction accuracy for better-informed capital investment/lease decisions on restaurants. Online platforms can also utilize our method for premium business intelligence. Furthermore, our findings regarding consumer-posted photos can provide foresight to managers and investors in terms of survival likelihood for up to three years, which can be highly valuable for competitive intelligence, resource allocation, and longer-term strategic planning.
References
Heyman, Stephen. 2015. “Photos, Photos Everywhere.” New York Times, July 29.
Lafontaine, Francine, Marek Zapletal, and Xu Zhang. 2018. “Brighter Prospects? Assessing the Franchise Advantage Using Census Data.” Journal of Economics & Management Strategy.
McGrath, Thomas. 2017. “Social Listening Meets Image Sharing: A Picture Is Worth 1,000 Words.” Marketing Science Institute.
Parsa, H. G., John T. Self, David Njite, and Tiffany King. 2005. “Why Restaurants Fail.” Cornell Hotel and Restaurant Administration Quarterly 46(3):304–22.