Current online video interaction is typically designed with a focus on straightforward distribution and passive consumption of individual videos. This "click play, sit back and watch" context is typical of videos for entertainment. However, there are many task scenarios that require active engagement and analysis of video content as a means to an end, such as educational material for flipped classrooms and critical review of performance — in sports, performing arts, or other spatial domains. Interface design has focused on the former context in interface interaction controls, and common interaction tools are limited to basic player affordances such as play/pause, speed change, and scrubbing. These basic interactions, however, can be cumbersome for goal-directed tasks such as contextualized communication, information finding, or comparative analysis. More active, goal-directed use of online video requires better interface designs to support these kinds of task-based interaction.In this research, we investigate approaches that enable and support more active user interaction with online video. In order to conduct the investigation, we present an illustrative research probe that employs analytics on user engagement to facilitate interaction. For example, video segments that are re-played by many users may indicate an important concept to review in a flipped classroom video. To study this kind of interaction, we introduce metrics for measuring the degree of user engagement and an interface component for visualizing and interacting with user engagement data as a heatmap. To evaluate our research probe, we have conducted a series of user studyexperiments. The set of experiments studied the impacts on user interaction of making engagement data available across multiple levels: individual user, small groups, and large groups such as a general population. The results of the studies show that the heatmap interaction approach can provide better support for active interactions with online videos such as information finding, content understanding, and comparative analysis. More specifically in the online learning context, our research prototype supports viewers to find information more effectively and also helps them feel more confident in their learning. For the group of class instructors, the study shows that our research prototype can better support the development and refinement of teaching methods and improve course video content.