Beyond Views: Measuring and Predicting Engagement in Online Videos
share
Online videos account for 73 and proportional amount of traffic are still growing. Hence understanding how videos capture attention on a global scale is also of growing importance. Most current research focus on modeling the number of views, but we argue that video engagement, or time spent watching is a more appropriate measure for resource allocation problems in attention, networking, and promotion activities. In this paper, we present a first large-scale measurement of video-level aggregate engagement from publicly available data streams, on a collection of 5.3 million YouTube videos across 2 months in 2016. We study a set of metrics including time and the average percentage of a video being watched. We define a new metric, relative engagement, that calibrates against video properties and strongly correlates with recognized notions of quality. Moreover, we find that engagement measures of videos are stable over time, thus separating the concerns for modeling engagement and those for popularity - the latter is known to be unstable over time and driven by external promotions. We also find engagement metrics are predictable from a cold-start setup, having most of the variance explained by metadata, content and channel reputations - R2=0.77. These observations show that engagement metrics, measured or predicted, are valuable tools for content producers and hosting sites to choose or rank content even before observing its audience reactions.
READ FULL TEXT