In a fairly recent article on iMedia, Brandt Dianow summises that Google Analytics is inherently inaccurate due to the way it counts ‘bounces’ as visits. While the article was very well written and well thought it, I just do not agree with the premise that a bounce is not a visit. For example, if I searched for something on the web and a Wikipedia page showed up on the SERP and I clicked on that link, read the definition and then bounced back to the SERP page is that not a visit? Can you imagine how this small difference in the definition of a visit would impact the stats on a site such as Wikipedia. In addition, how could you possibly not define that as a visit and a successful one at that. I found the information I was looking for, read it, and bounced!
The problem, as I understand it, with counting a bouncer as visitor is that Google Analytics and many other page tagging based analytics packages do not have way to calculate duration for a bounced visitor. Because of this, all bounces are counted as zero duration visitors which I’ll agree is inaccurate, but still does not change the fact a successful visit could have been acheived. In the comments on Brandt’s article someone mentioned using the PageAlizer tool which supposedly is able to calculate duration for a bounce.
Rather than toss the baby out with the bathwater I would recommend a more hybrid approach to solving this ‘issue’ in Google Analytics. Perhaps a setting is in order. This way a user could set their preference on what to count or not count as a visitor in relation to a bounce. I would be happy with that how about you Brandt?
I did not say this was a “stop” problem – the focus of the article was to provide a way of getting a more accurate measure of average duration, via a simple formula I developed which removed the 0-duration visits from the calculation. You can also used Google’s segmentation to create a better metric.