Recommendation technology is often applied on experiences consumed through a personal device such as a smartphone or a laptop, or through personal accounts such as one’s social network account. However, in other cases, recommendation technology is applied in settings where multiple users share the application or device. Examples include game consoles or high-end smart TVs in household living rooms, family accounts in VOD subscription services, and shared desktops in homes. These multi-user cases represent a challenge to recommender systems, as recommendations exposed to one user may actually be more suitable for another user. This talk tackles the shared device recommendation problem by applying context to implicitly disambiguate the user (or users) that are being recommended to. Specifically, we address the household smart-TV situation and introduce the WatchItNext problem, which — given a device — taps the currently watched show as well as the time of day as context for recommending what to watch next. Implicitly, the context serves to disambiguate the current viewers of the device and enables the algorithm to recommend significantly more relevant watching options than those output by state-of-the-art non-contextual recommenders. Our experiments, which processed 4-months long viewing histories of over 350,000 devices, validate the importance and effectiveness of contextual recommendation in shared device settings.