Despite the increasing interest and need of marketers for planning effective social media engagement, there has been relatively limited academic attention on examining the role of the when and what elements of social media marketing. This study proposes a modeling framework which can address the role of the time stamp and media type in social media engagement. Based on the elaboration likelihood model (Petty & Cacioppo, 1986) for information processing, we conjecture that media type and time-of-day will simultaneously matter in social media engagement. By text-mining Twitter data for a new brand in the music industry, we empirically investigate how consumer preferences for different media types (e.g., video, picture, and online news) can vary across the time of day. We find that unobserved consumer heterogeneity for media type preferences should be incorporated to simultaneously examine the when and what elements in consumers’ social media engagement. As our data is from the music industry, non-text based media types are generally preferred over text-based media types. Among non-text based media types, video is preferred over picture in the morning, whereas picture is preferred over video in the daytime. Moreover, text-based media types are preferred early in the day compared to later in the day.