We will refine this in the new 8.100 reporting a bit, but basically time is always going to be an educated guess, there is simply no way to 100% know (without a very invasive type of client) a site which is open vs. being read vs. actually having 'time spent' at it. We see the requests made when a user makes a click.
Add into this multiple tabs, and various habits as to how people surf (say doctors vs. high school students) and it becomes a good challenge.
I'm paraphrasing as i dont have it all right in front of me, but loosely, we watch for the request and start a time counter, requests made in the same 5 minute window for that site refresh the counter again. When we don't see a request to that site for 60 seconds, we deduct 4 minutes. This is based on an assumption that most people wont read a single page for longer than 60 seconds without somehow interacting with it again. Over this, they have probably left and we treat it as a new hit.
I'm paraphrasing as i dont have it all right in front of me, but loosely, we watch for the request and start a time counter, requests made in the same 5 minute window for that site refresh the counter again. When we don't see a request to that site for 60 seconds, we deduct 4 minutes. This is based on an assumption that most people wont read a single page for longer than 60 seconds without somehow interacting with it again. Over this, they have probably left and we treat it as a new hit.
Thank you Angelo!
Obviously, in ajax web sites such as facebook (in which there is a continuous communication between browser and server) if you left the client open the time counter runs continuously, is it right?
