Too many website owners are tripped up on the false notion the analytics measurement of “bounce rate” is their main enemy.
Now this may get a bit complicated for those who aren’t new media douchebags like me. But it’s important and interesting (at least to me.) But before I get too geeky, let’s back up.
What’s a bounce rate?
There are several “exact” definitions of bounce rate around the web; here’s mine:
Bounce rate is the calculation of visits to a website requesting only one page call to the server DIVIDED BY the total number of visits resulting in multiple calls.
But what happens when a website can serve up multiple PAGES during one PAGE CALL?
An example of this is the site we recently just launched for a client specializing in high-end loft leasing.
This is a very bouncy site.
Of the five content areas (pages), all are loaded at once. This enables the fancy scrolling function afforded by the navigation, and also reduces the time it takes the viewer to receive additional information on the property when going further than the home “page”. Residential leasing in major markets is a crowded web category, so all the frustrations we can reduce for the browsing shopper, the better. Even if the saved time result is in milliseconds, the effect of the navigation is pleasing and can result in some interesting design interactions.
However, one of the consequences of the act of loading all five pages from the outset is that it produces a completely irrelevant bounce rate metric. And we knew that going into the build. For our needs, the % of new visits, average time on site, and server logs showing total numbers of .pdf dowloads (floorplans and applications) give us a much better, and much more accurate, description of this website’s effectiveness.
Which brings me to another point—even for a traditional page loading execution, the bounce rate metric is overrated. Not waaaaaay overrated, just over-emphasized, over-scrutinized, and taken as being an absolute representation that it absolutely is not.
I’m not being semantic; “bounce rate” is an incorrect term. “Bounce” is an inference; an assumption of users’ behavior based on data (data it should be noted, that has a not-insignificant margin of error.) To that end, “average time on site” can be problematic, as well, since it would be better called average time user left browser window open, with an eventual timeout threshold (and of which, it is unclear on how said timeout affects average time calculation).
New web masters (of which there are many thanks to the explosion of wonderfully simple new content management systems) can’t be faulted for putting hefty stock in these data calculations. People have been looking at website server logs for a long time trying to figure out how to interpret the data; Goggle, with their analytics platform estimated to pervade over 80% of all sites in the world, just came up with a really slick way to package and present that data. Their contribution to data visualization cannot be understated.
Google Analytics spawned inferencing of your users’ behavior and attitudes, though analytical, is still best left as a theory-based practice for most web properties. No number is absolute, and no number should be taken on its own without further cross reference.
Google Analytics is an extremely powerful tool for reporting and monitoring incoming and outgoing website traffic. But the power of its data is relative, and each website needs individual evaluation based on the owner’s needs, wants, the site’s architecture, the industry category, and main marketing methods to determine which, of all possible analytic measurements, are most important to evaluating website performance.
For example, relevant web traffic is much more than the opposite of whatever your bounce rate might be. The analysis of analytics is still an art form, and ultimately, requires an opinionated read of the data to be truly useful. When it comes to web server logs, data is very much our friend, but as is the case of Google Analytics (like other measurement platforms), the big picture created by each data relationship is much more important their respective quantitative calculations.
Hey look, you’re still here! Maybe this was interesting…
Just so you know, google analytics which the example in your site uses, can actually track clicks on the navigation on the left hand side of the page.
You must use custom event tracking. For more information about it go here: http://code.google.com/apis/analytics/docs/tracking/eventTrackerGuide.html
This will let you know if users are interacting with elements of your site that do not change the url. Event tracking creates a virtual page view, that will give you a better idea of your real bounce rate.
So don’t be to harsh on bounce rate, all analytics must be looked at in context to gain true insight, and I completely agree, you should never look at only one metric to make decisions.
P.S. you current site could use event tracking as well for better analytics, however, you would have to change to the new google analytics tracking code.
What’s up Tim, thanks for the comment.
Yes, custom event tracking is pretty cool, but it too just tells part of the story. That data may suggest how many users interact with the left side links to view the page content, but it still leaves out info on who might scroll to it on their own.
In that sense, it’s still pretty faulty data. We wouldn’t accurately be able to say XX% of users view the content page based on the custom event tracking. In the unlikely event that the sum of all custom events equals the total number of visitors, then we might be in business.
But until then, it’s inference leaned atop inference, and that’s not something Google Analytics leads on too much about.
Sometimes we get phone calls from people that find us online and want to know how much we’d charge to do their nails. I kinda wish those people would have bounced.
Irrelevant traffic doesn’t really hurt your site but it jacks up the bounce rate. This can sometimes be a tough thing to explain to clients. It’s almost as if bounce rate is a better measure of how well google is doing.