In my experience, time on site isn't going to be a useful metric however you go about it. At least with the news sites my company runs, checking if a page is still open would be hopeless because it's very common for users to just open the 2, 5, 20 pages they want to read in different tabs, then work through them slowly. And for clicking outbound links, well nearly all exits for us are window/tab closing, not links elsewhere.

Maybe this isn't so much the case for other industries, or even for the same type of site targeting a less techy audience? (e.g. My mother wouldn't ever open multiple tabs as a queue of pages - but then, she might well open a page and then go do some cooking, or whatever, before reading it.)

If google is using bounce rate as a signal for page ranking, this advice is very useful for people who have a low bounce rate simply because all the information is on a single page, or because the page uses dynamic html to refresh the content as the user digs deeper into the app.

neat. I wish I had a bit more time to play with your TOS gist. Unfortunately, I just skipped my morning workout to get what I got in the comment there, and now have to run off to work. (boo)

Oh well... that's why they call it a gist right? :) I think focus and blur events on the document or window are key to getting "perfect" TOS. Just have to think about how to make it most efficient ... I prefer doing away with the timer and just sending the data once in the UNLOAD event.

There's peril to that (browser crashes?) and maybe the 10 second 'tick' is better.

Dunno. I really have to go now. Breakfast! Starving.

Another way to do this is to trigger the tracking with the window.onbeforeunload event paired with logic that makes sure the user has actually been on page for longer than X seconds/minutes.

I always supposed that G was doing that anyway. Why would they measure time-on-site only per page load?

Has anybody implemented this and seen a jump in their time-on-site and bounce rate values?