|Estimating and Optimizing the Impact of Photo Assortment in Sharing Economy
|Year of Publication
|Li H, Simchi-Levi D, Wu MXiao, Zhu W
Problem Definition: We study how listing images displayed on Airbnb affect customers’ booking decisions, and we suggest optimal photo assortment to increase listings demand. Academic/Practical Relevance: Host-generated property images as a visual channel reveal substanstial information about the property. Selecting proper images to display can lead to higher demand and increased rental revenue. We define, estimate and optimize the impact of photo assortment in the setting of Airbnb. Methodology: We apply Resnet50, a widely-used convolutional neural network model, to build two separate supervised learning models to evaluate image quality and content posted by Airbnb hosts. Then, we construct a novel pairwise comparison model to consistently estimate the impact of photo assortment on listings’ attractiveness. Furthermore, utilizing our estimation results, we build a non-linear integer programming optimization framework and develop an algorithm to derive optimal photo assortment for each listing and calculate the revenue gain. Results: Our estimation results suggest that high quality bedroom cover images lead to the largest increase in listings’ attractiveness. Also, posting photos featuring the same room type leads to discounted marginal return. Finally, our counterfactual result shows that the optimal photo assortment suggested from our algorithm leads to 14% increase in rental revenue. Managerial Implications: Rearrangement of displayed photos can unlock tremendous value for the listing owners and improve listings demand.