Forum Moderators: DixonJones
Presumably GA actually has no way of telling when a user leaves a site. All it knows is if/when the user clicks to another page on that site. I'm assuming that all it does is lays a cookie and tracks if that same person requests another page from the site. It doesn't seem to do anything more clever like using a javascript onunload event, for example. So it's pretty blind.
But what if a user comes directly to a page on my site to read a long tutorial, which is contained on a single page. They might sit there and go through the tutorial for 15 minutes, and find it excellent. Then they might leave and go to another site. If I'm right about my above assumptions, then GA will record this as a 'bounce', because they only viewed a single page.
And how will it give any indication at all about how long the user was on the site? All it can do is assume that a user is clicking other pages on the same site while on a 'visit', and record the time between the first and last page request, then divide that by the number of unique IP addresses, to give average time on site. So the visit from fictional user reading my tutorial will actually have a negative effect on the 'average time on site' figure, even though in reality he spent 15 minutes reading it.
Where I work, we modified the data collection tag (not Google Analytics, actually WebTrends) to detect the first hit of a visit and to send a signal every x seconds, up to a top limit that we specify. With a little judicious filtering in the reports, we then get information on how long those bouncers actually stayed. It's extremely helpful.
I first saw this method on a site called The WebTrends Outsider (they provide some code). It might be able to be adapted to Google Analytics, but I'm not sure that Google Analytics reporting will be able to report properly on it. The "other" analytics program has a lot of filtering, measuring, and manipulation capability that I can't seem to get in GA, though maybe it's just ignorance.
im looking to get something more than than the reports GA is providing..i looked at Clicktale and Click Density...
are you using Clicktale? what kind of strain does it put on your server if you are using it?
im noticing that google site overlay is worthless and does not track any clicks from my CSS menu drop downs....
so im considering integrating a more "user tracking" based application.
There are couple of things that perhaps they could give you a more accurate picture. If you have a web store selling products, setup the shopping cart page as part of your GA only when the cart contains products. Then based on that calculate the bounce rate. Usually adding products to the shopping cart requires use of forms that can be easily modified to be secure enough and filter out bogus traffic.
Personally I do not rely on GA or other jscript driven statistical packages mentioned here, but instead measure the real traffic from the server end as I tend to support as many different browser settings as possible to maximize sales. Therefore if some of my customers are browsing the site with active content off, I want to measure that too.
We haven't noticed any page load slowdowns as far as the JS loading off their servers.
