Question

I have access to a proxy server and I can find out the time a video was requested. The log has the form (time, IP, URL). I want somehow figure out for how many seconds did a particular user using IP address A watched a YouTube video. Any suggestions?

Answer 1

If you only have access to requests, you obviously can’t tell the difference if someone just loaded a video or watched it.

So, the best you can do is to come up with a set of heuristics that tries to ‘guess’ it by observing certain actions of the user. Here are a few ideas:

• Does you log count the requests for the video buffer itself? If it does, you can see how much of the video was actually loaded, and the watched time can’t be more than that.
• If you (quite naively, I guess) assume that they’re finished watching when they request another video URL, you can use this as your trigger for ending a ‘video session’.
• Install Wireshark or similar and start watching activity from YouTube during the video. Can you identify if there’s a request when advertising is shown, or the related videos are displayed when the video finishes?

In all honesty, though, I think it will be virtually impossible trying to derive such an specific metric like seconds watched from such limited data as the point in time a video was requested. Just think of what could mess up any strategy you come up with: the user could load several videos in different tabs in a burst, or he could load a video page, pause it and forget it for several minutes or hours before he does watch it.

In short: I don’t think you’ll get a reliable guess using only the data you have, but if you absolutely must at least try, observing network activity between client and YouTube that only happens when a video is in the ‘playing state’ (pulling advertisings, related videos, some sort of internal YouTube logging, etc) is probably your best bet. Even that probably won’t have a granularity nearly close to seconds, though.