Why Average Time On Site is a Bad Metric
Average Time On Site is a flawed metric as analytic packages really aren’t able to track the time spent on any given page of your site. But wait a minute, don’t all analytic packages provide an average time on site metric? How do they do it? First, they calculate a visitors time on a given page as the difference between when they go to the page, then go to another page within the same site. Then they add all of that visitor’s time on page calculations together to get their time on site. This seems reasonable enough on the surface, but has three main problems.
Problems with Average Time On Site/Page:
- Time on page can’t be measured if a visitor only views one page of your site (i.e. they bounce). That doesn’t mean that they didn’t spend time reading your content and possibly performing a valuable action such as bookmarking your page or forwarding it to a friend, but the amount of time that they’re on your site cannot be calculated as they didn’t click on another page of your site.
- Time on page can’t be measured for the last page that a person views on your site so it’s usually counted as zero seconds. This is true for everyone that visits your site. That’s right, EVERYONE. And that last page may be key–it may be where they spent the most time.
- It’s not often that you have 100% of your visitor’s attention, so their time on your site is often interrupted, interspersed with side tracks to other sites (some of which you encourage by linking out to other sites from your site), phone calls, coffee breaks, or a quick trip to the bathroom. In fact, I know many people who open a page of interest with no intention of reading it right away and then leave it in an open tab until they have time to get to it (sorry to break it to you, but that person who spent 36 hours pouring over our latest blog post, probably didn’t). You get the idea. Any one of these activities will inflate your time on site number!
On top of the inherent flaws in calculating this stat, analytic packages further muddy the waters by making different assumptions about how to handle these situations. Should a person who was on a page for 36 hours be counted as a 36 hour time on site, 20 minutes, or zero seconds? It depends on the analytic package you’re using! Does that visitor who bounced count as zero second or some estimated amount of time? Again, it depends on the analytic package you’re using.
Further reading: The following blog post provides a very in-depth explanation of how average time on site is calculated and examples of many scenarios where it provides misleading numbers… http://www.searchenginejournal.com/tick-tock-the-limitations-of-average-time-on-page-and-average-time-on-site-in-google-analytics-experiment/21439/
What does all this mean? In a nutshell, the average time on site metric is not to be trusted!
What should you track instead?
As a bare-bones measure of engagement, average page views is better as at least it’s accurate and consistent between different analytic packages. A page view on the last page is counted the same as a page view on a bounced visits, which is counted the same as every other page view, so at least it’s consistently and reliably calculated.
But better yet, if you’re really trying to get a measure of engagement, then tracking micro conversions will give you a better picture of what your site’s visitors are really up to. Micro conversions are just actions (tracked as events in Google Analytics) that you track within each page of your site, such as the number of people who wrote a product review, commented on your blog, shared your page to Facebook or Twitter, or interacted with your online chat. With some javascript programming, you can even fire off events when visitors scroll below a certain point on the page (tracking how many visitors view below the fold) or when controls on embedded videos are pressed (allowing you to track how many visitors not only start viewing your videos, but finish them).
Knowing how people are interacting with your site will provide much deeper insight into what’s working and what’s not than average time on site ever could.