Time on Site & Time on Page – Google Analytics metric mystery
When asked whether ‘time spent on site or page’ hold any meaning in a website’s performance, I have always said that, “if it means anything to you”. But before we can relate any performance measure to them, we need to understand how are they calculated and how relevant are these two metrics.
How is Time on Page and Time on Site Calculated?
Time spent on site is calculated by generating time stamps on visit to every page and then calculating the difference between the last and first time-stamp of a visitor session. For time spent on page time stamp on entry to that page and time stamp on entry to next page are taken into account.
Let us consider the diagram below:
The above diagram should make it clear how time stamps are used to calculate Average Time on Site and Average Time on Page. It should be noted that since the visitor exited from Page 3, therefore, there is no succeeding time stamp and hence time spent on Page 3 cannot be calculated and is shown to be 00:00 by Google Analytics.
So this means that if a page view of a particular page leads to an exit, then the time spent on that page is shown to be 00:00. In case of a bounce, time on site and time on page are both reported as 00:00 minutes. What is Exit Rate and Bounce Rate?
How reliable is this method of recording Time on Page & Time on Site?
Since time on page cannot be fetched for bounces and exits, the reported values are not 100% accurate because the data for exited and bounced traffic on a page is missing.
Reliability of Time on Page = 100% – Exit Rate
The above is true because Exit Rate indicates the percentage (number) of visits that led to an exit from that page so for that many percent of visits the data is missing and hence not reported. However, the Average Time on Page is calculated by taking 100% of page views (visits). So if my a web page has an exit rate of 20% then the reported Avg. time on page is only 80% accurate.
Lower the Exit Rate more reliable the Avg. Time on Page.
Actual Time on Site = Reported Time on Site x 100 / (100 – Bounce rate)
The above equation discounts the contribution of bounces. A bounce contributes 00:00 and hence causes an error because a bounce is taken as a page view (visit) whatsoever while averaging the Time on Site. Bounces should be left out because what we are are more concerned about is the visitors who browsed through the site. So, if my site has a Bounce Rate of 30% and the Reported Time on Site is 120 seconds (00:2:00 minutes) then the actual time on site is
120 sec x 100 / 70 = 171 sec = 2 min 51 sec
Note: This method assumes that reported Avg. Time on Page for all pages is correct.
Lower the Bounce Rate more reliable is the Time on Site.
And if you are to approximate the Avg. Time Spent on a Group of pages (a section of your site) then you can replace the bounce rate with Avg. Exit rate for those many pages in the second equation.
This method of calculating Time on Site is also not 100% accurate because even the reported time on site is faulty to begin with. Time on Site is nothing but the sum of Avg. Time on Page for all web pages. And we know that the reported Time on Page is not 100% accurate because Google Analytics does not record time spent on page for exits and bounces.
While you’ll be reading this I’ll be preparing a tutorial post on How To Calculate Time on Site with Maximum Accuracy. Until then enjoy this curry!
No related posts.
Comments
9 Responses to “Time on Site & Time on Page – Google Analytics metric mystery”
Feel free to leave a comment...
and oh, if you want a pic to show with your comment, go get a gravatar!
All site analytics show you some stats on how long people visit your page. Most site analytic services guess how long a visitor has been on a page by noting when a visitor arrives and that visitor views another page in the site. This calculation is a good estimate in general but has its weakness when calculating first page visit length – sometimes there isn’t a second page view. Pagealizer does actual counting. Our tracking code pings the tracking server as long as the visitor has the page open.
This metric is very important for landing/sales pages which a lot of times do not trigger another page view and help you understand if people read your page. If for example you have long copy and most people visit your page for 10 seconds you might need to add more call to action links on the top of your page or reduce the amount of text. Check us out
Ah, this is a GREAT article. All of the 0:00 page views in google analytics were confusing the heck out of me! I thought they might be from automated bots? (is there any way to estimate that?)
I took your formula and applied it – very helpful.
personally i find it a little easier to compute this more quickly by multiplying the reported site time by 100/bounce rate.
example:
*reported site time = 65 seconds. *bounce rate of 50%
**actual time: then 100/50% = 2 and 2*65 =130seconds.
Re: eye tracking research
Yes the formula you applied gave you the actual time spent on your site i.e. the time spent by actual visitors who stayed on your website. This formula excludes the visitors who bounced without navigating further into your site.
Hi Abhi this is great article, but can you elaborate on Time on Page & Time on Site, I mean according to your article are these 2 totals should be equal? If No then can you please explain why?
this post is incredible helpful! i could not figure out what all those 0:00 visits meant and your site has the clearest explanation.
so another question – is there an easy way to remove or filter these “bounce” visits from our site analytics? it seems like these visits from bots or other sources are skewing our site metrics – is that correct? is there a standard practice around this that you would suggest?
thanks
[...] Which is why the emphasis on collecting stats from at last two pages: given the current crop of analytics tools that struggle to do anything meaningful with single page visits, specifying a two page visit means that not only visits to the site that are likely to be meaningful are reported, but also that the reports are more likely to contain meaningful data too. (There is an obvious problem here: if visitors visit two pages, and quickly click to the second from the first before exiting the site from the second page, the time spent on the second page won’t be captured? See for example Time on Site & Time on Page – Google Analytics metric mystery) [...]
[...] Which is why the emphasis on collecting stats from at last two pages: given the current crop of analytics tools that struggle to do anything meaningful with single page visits, specifying a two page visit means that not only visits to the site that are likely to be meaningful are reported, but also that the reports are more likely to contain meaningful data too. (There is an obvious problem here: if visitors visit two pages, and quickly click to the second from the first before exiting the site from the second page, the time spent on the second page won’t be captured? See for example Time on Site & Time on Page – Google Analytics metric mystery) [...]
That was a highly informative article…Thanks for a lot for posting it…keep up the gr8 work…
Regarding the proposed adjustment of screening out bounces, that will help, but it still does not provide an accurate estimate because a visitor could land on your page, click off to a second page after 5 seconds, and then stay on that second page for 20 minutes, but the time on site would still display as 5 seconds.
One of my sites has a curious variant of this problem because it loads an iframe containing help pages on the main page using Javascript a few seconds after it comes up, and so the report has numerous visits that are a few seconds long, and no bounces at all.
My main page has interactive Javascript content, and that content is the main reason a visitor would come to the site, so it is entirely possible for a visitor to spend the entire visit there without loading another page.
A real solution might be for GA to hook the browser events in Javascript to detect user activity and then when a browser close was finally detected to report the time between the first page load and the most recent user action (not including the close). This could also solve the problem of users leaving the browser open and running off to do something else. Maybe GA does not want to do something that invasive and potentially disruptive, though, so it is what it is.