A reliable buffered approach for measuring usage of web video

A popular way to track video usage is by sending periodic heartbeats from the player. In a given video session, heartbeats identify that the session is alive, but do not capture the user´s actions on the video (play, pause, seek, etc.). If the video providers use different frequencies for the heartbeat, usage data for the same video published to different websites cannot be compared across publishers in a standardized way. Also, the number of heartbeats sent for each video increases the cost of tracking. In this paper we propose a reliable way to capture video usage data, where we buffer all the user actions, and send the buffered data at the end of the session using a single server call to the tracking server. The entire time-line of the session can be reconstructed from the collected data for a fine grained view of the video session. However, a user can quit the webpage abruptly by navigating to a different website or by closing the browser (or the tab). In the event of such abrupt session ends, the browser application may quit before sending the buffered data successfully to the tracking server. We delay the browser application closing and optimize it to be within an acceptable user experience. We base its value on the round trip time (RTT) between the client and the tracking server. We measure the RTT at the beginning of a session and derive the requisite delay for the session as a function of the RTT. We implement the delay as a busy routine in javascript and test across all browsers. We arrive at the optimal value of the delay with simulations in various adverse network conditions. We also experiment by having the client and tracking server in different geographies. Finally, we validate our method in a popular website accessed by heterogeneous clients with encouraging results.