Are you a Google Analytics enthusiast?
More SEO Content
Sudden Drop In Bounce Rate And How To Track The Cause
Posted 05 February 2010 - 05:02 AM
It's now around 15 days since the bounce rate dropped and it's still down, but I can't seem to find a corresponding change in traffic sources or any other cause - any ideas how to track it down?
Posted 05 February 2010 - 11:24 AM
Do you do anything with Segmentation by chance Rolf?
The reason I ask is there was a change in how Google handles the SetVar in late January and I'm wondering if this change may be responsible for what you're seeing.
What they did with the change really should have a different and good effect, since the way it was set up before the SetVar would fire on a page and be recorded as a 2nd interactive hit, thus making the first hit (to the page itself) a 0 second page view. The way it was before really messed up the Time On Page and # of Page Views for those who were using Segmentation in any sort of large scale.
I can't quite wrap my head around how the change would affect bounce rate the way yours has changed. Logically it should increase it, not decrease it. But as we all know when you make a blanket change like this it can certainly have unintended effects. And the timing is dubious since you're change in BR happened in January and they also changed the SetVar routine in January.
Any idea what day the bounce rate started to skew? Knowing that may help us track down a cause if it's on GoAn's end.
And as a FTR... If anyone reading this is using GoAn with out-of-the-box settings the SetVar change should have no effect on your stats. In theory anyway. It should only affect those who have Segmentation set up. Again in theory.
Other totally unrelated things to look for when you see a drastic, unexplained drop in Bounce Rate include:
- Duplicate Tracking Code - If your pages somehow (usually via an include of some type) place the GoAn tracking code on the page twice, it ends up recording each page view as 2 page views. Thus causing Bounce Rate to drop off a cliff. The easiest way to see if this is happening view the source code from your browser and search for the ga.js or urchin.js references. If there's more than one there in the source code, you have this duplicate tracking code problem.
- Frames/iFrames - Same sort of situation as above, but this time the main page that contains the Frame/iFrame includes tracking code and the Frame/iFrame also includes the tracking snippet.
- Ajax Routines - This one happens can cause problems if you have your Ajax routine set to send an update in the trackpageview module when the ajax reaches document ready status. It's not necessarily a bad thing to have the Ajax send this trackpageview notation since for the user they are actually seeing new content. The trick is to make sure trackpageview is not being sent when Ajax fires off something that doesn't require a user click. In other words, your initial page load and initial default Ajax routine to produce the first page's content should not update trackpageview. It should only be sent after there's been actual user interaction.
- 3rd party vendor stuff - This is very, very similar to the duplicate tracking code situation, but is one where you have less control to fix it. The issue is one where if your page uses GoAn to track and the other page also uses GoAn for tracking, presumably under a different UA account number, it can still mess up bounce rates.
Posted 08 February 2010 - 04:32 PM
I knew from the way the pages were set up there was no frames/iframes or 3rd party stuff to deal with and I was fairly sure all the AJAX was kept well away from the GoAn stuff, so I looked for duplicate js to see if it was registering 2 hits for each page load.
That didn't lead to anything directly either, but as I was examining the code it hit me like a plank of wood round the head that I had updated the GA code from an old version to the latest version in January, and sure enough the dates matched up, so I suspect that is the cause.
I still can't explain why it's had the effect on bounce rate, or be certain about which rate is the correct one, but it's given me more to investigate. Any further suggestions are welcome and I'll post the results if I find anything that might be useful to other GoAn users.
Posted 09 February 2010 - 03:15 PM
Feel like doing a bit more investigation?
If it were me I'd do a quick search of both google.com and of the Google Groups area of GoAn to see if you can find anyone else reporting something similar happening when they switched to the new style of code. Then when you don't, and I suspect you won't , I'd try to get in touch with Avinash Kaushik to see if he can help you track down the cause. As well as which numbers are more correct.
Avinash is really good about helping out in these types of situations. And knows more about the back end processes in GoAn than anyone else I know. Seriously, he's said in the past that he gets roughly 50 questions per day through his blog, and it's one of the joys of his job to answer those 50 requests per day. I think this will be one that will interest him and may even turn into a blog post for him when he tracks down what is happening. Hey, maybe you'll even get a bit of free publicity out of it for your site if that happens.
Rather than post it here, if you scroll down a bit on the About page of his blog you'll see an email address you can use to get in touch with him. It's not in a normal email addy snytax, so look for the part that says blog at.
I'd be really interested in finding out what he has to say about the matter. I think it'll be a good one to file away for the future since it sounds like you've tracked it back to a pretty high degree of certainty that what you're seeing in bounce rate is attributable to changing from the old style to new style GoAn snippet.
Posted 09 February 2010 - 04:35 PM
From what I am seeing, it looks like the bounce rates have stayed about the same over a large period.
Not saying this proves anything, but it gives a *possible* comparison.
If any specific filters/segments/etc would help out, let me know and I'd be happy to check it out.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users