A few months ago I started an experiment. I added some code to my site to better track how visitors interacted with my content. The new code tracks various user activities, like scrolling and reaching certain locations in the page. Not only did the code collect more data about user behavior it also changed many established metrics, like time on site & page and bounce rate.
The result is an entirely different view of a user visit and new benchmark metrics for blog performance.
At least that’s my opinion. I’d like to hear yours :)
You can read more about the code in these articles:
Advanced Content Tracking with Google Analytics: Part 1
Advanced Content Tracking with Google Analytics: Part 2
Time & Bounce Rate
The main reason I added the code was to get a more accurate measurement of the time people spend reading the content on the blog and the bounce rate.
Before implementing the code the average visit length for my blog was 1:40 (one minute and forty seconds) and the average bounce rate for the blog was 78.34%. This was fairly consistent with what I had seen for other blogs.
But after adding the code the average time on site for the blog adjusted to 9:34 (nine minutes and 34 seconds) and the bounce rate for the blog fell to 23.11%.
Better metrics for a blog!
Over the last 4 months these metrics have averaged 8:30 (eight minutes thirty seconds) for time on site and 22% for bounce rate.
The reason the metrics changed so dramatically was due to how Google Analytics calculates time and bounce rate. By adding more information about on-page behavior, Google Analytics had more data point to calculate time.
You can read more about how Google Analytics calculate time.
Let’s dig deeper.
Averages can be easily skewed, so I often use the Audience > Behavior > Engagement report to view a distribution of visit length. After the code change this showed some interesting results.
Using a distribution of time we see that the 0-10 second visits decreased and there was a ‘flattening’ in the time distribution. Again, more accurate measurement of people reading.
After adjusting the code the time calculations in Google Analytics became much more accurate.
Moving beyond some of the standard metrics and reports, you can use the Event Flow to visualize how people move through the content, scrolling and reading articles. I wrote about this when the Event Flow reports launched.
I really like this view. What’s particularly interesting to me is that most people leave the site when they get to the bottom of the content. They don’t read the comments and only a few actually click through to another article.
Only a small number of people read a second blog post. Even LESS read the blog comments.
All of these metrics are available in tabular form as well in the Event reports. I don’t use these reports too much, instead I’ve created a number of goals to track micro conversions on the site.
I track how many people get to the bottom of the content and how many get to the bottom of the page.
Tracking these actions as micro conversions makes it easier to segment the data by traffic source and other dimensions.
This is particularly useful when segmenting by mobile. Mobile visitors don’t reach the bottom of my articles as much as non-mobile readers. 48% of desktop users read to the bottom of an article while 32% of mobile readers reach the bottom.
Using a goal makes it easier to measure those getting to the bottom of content and the bottom of the page.
As I get ready to redesign the site I’m going to dig deeper into mobile segments and review screen sizes and how reading behavior changes for those on different devices.
Wrapping up….
Did changing this code revolutionize how I track my site? No, not really. But it gave me more confidence in the data and painted a more accurate picture of what happens during a visit.
Am I glad I did this? Absolutely. I have more confidence in the data and a much better understanding of my users.
Should you do this? That depends. It could be a big challenge for your IT team. You need to decide how important accurate data is and prioritize this work.
BUT, the next time someone asks you the average time on site for a blog you can say 8:30 (eight minutes thirty seconds) the bounce rate should be 22%.
Great article. I have long told certain clients to not place much stock in bounce rate, for example when the CTA at the bottom of a landing page deliberately drives visitors off-site.
Hi Justin- Well, I was one of those people who read both articles..a few different times. I liked it so much that I presented to a few product divisions and we are testing a version of it on one of our knowledge bases. We are scheduled to roll it out in two of our other content sites.
The thing we have been thinking about is having bands instead of just Readers and Scanners. So, 1 to 3 minutes’ 3 to 5 minutes; 5 and up..not exactly, just citing an example. Our first execution of it is on an online knowledge base and the time they spend reading articles is very high because its very useful, detailed information on how to code and configure things.
I have had a few training sessions so far with the team who is using it but we need to get together again and focus just on that. We also use other custom visitor segments and I have get them fired up about playing in the data.
Anyway, thanks, Mary Kay
Nice article Justin. What do you think of the adjusted bouncerate for Google Analytics? Is that sufficient to track your real bouncerate?
@Gerard: I don’t like the adjust bounce rate method. There’s nothing there that really implies engagement. I would prefer to tie the event to some type of action, like I do in the Advanced Content Tracking method.