Dobrian et al. present a prime example of thorough, concise, well-thought-out research on a very relevant technology. The article reports on a study of the relationship between user engagement and video quality for video-on-demand applications. The introduction clearly and strongly argues the need for such a study. The researchers collected a substantial amount of data (user watch time, buffering time, pause time, and so on) from real-world applications, accumulating hundreds of millions of datasets, to analyze how user engagement is correlated with video quality. Perhaps the greatest contributions of the study lie in the collection, organization, and study of such a large dataset, and the fact that the data comes from real application users (instead of a small set of users in controlled laboratory conditions). The results reveal how user engagement increases with a decrease in buffer ratio (that is, shorter buffering time) and an increase in video quality. For live content, users are even more sensitive to buffering time.

The novelty of the approach lies in its quantitative measurement of both video quality and user engagement. Buffering time, join time, and paused instances are used as quantitative metrics for video quality. As metrics for user engagement, the study measures video clip play duration and number of plays (instead of asking for subjective user opinion through questionnaires).

The limitations of the study mainly pertain to the application area of video on demand (as opposed to, for example, videoconferencing) and the omission of classifications for content types (sports clips, movie trailers, news segments, TV episodes, and so on). The clarity, organization, and writing style make the article easy to read and understand for a wide range of audiences. As a result, it can serve as the perfect template for future studies on the relationship between video quality and user engagement. The article is highly recommended to anyone interested in measuring the impact of video quality in video-on-demand applications.