The short-term room rental online platforms have boomed in the era of digital business acceleration, as they offer excellent opportunities for interactive behavior between host and guests. However, pursuing customer repeated bookings is a challenge on these platforms. How to increase customer repeated bookings in the short-term room rental market? We examine theoretically and empirically this crucial research question. Drawing on the theoretical foundations of the social interaction theory, we identify and study three host-guest interactions that can affect customer repeated bookings: overall interaction, individual interaction, and offline interaction. We conduct a combination of empirical methods, including text mining and econometrics, using longitudinal data of 36,072 order records and review contents from 5002 listings and 2181 hosts in six cities Xiaozhu.com, which is a prevalent short-term room rental services provider in China. We find that overall host-guest interaction (online reply rate, acceptance rate, confirmation time) and individual host-guest interaction (the existence of individual interaction, absolute text length, and text similarity) increase customer repeated bookings. We also discover that offline host-guest interaction is a substitute for the effect of individual interaction (existence, text length, and similarity) on customer repeated bookings.