Tick Tock: The Limitations of Average Time on Page and Average Time on Site in Google Analytics [EXPERIMENT]

I’ve been told that I don’t have a very good poker face, and I’m ok with that.  I find it actually helps my clients, since they can tell almost immediately when I don’t agree with something or when a point needs more clarification.  For example, there are times I see marketers heavily focus on time-based metrics like Time on Page and Time on Site (as if those metrics were 100% accurate).  Unfortunately, this isn’t the case and the numbers reported can be far from accurate.

The confusion about how time metrics are calculated usually leads to some interesting conversations (as clients see various times listed in their reporting and want to start comparing pages and visitor segments.)  That’s hard to do, since there are serious limitations with the way time metrics are calculated in web analytics (for now anyway).  Even worse, people can subsequently make decisions based on the data.  This is why I tend to focus on key actions visitors can take versus other measurements of success.  I track these actions via conversion goals and events versus focusing heavily on time-based metrics.  I’m not saying that time-based metrics are useless, but they need to be taken with a grain of salt (at least until we find a more accurate way to track them).

For example, let’s say a marketing manager analyzed a new piece of content and was blown away with an Avg. Time on Page of 5:26.  This is where my (lack of a) poker face leads to a deeper conversation about how the actual Time on Page could be much different.  My comment is typically something like, “I think it’s great that visitors who decided to visit another page on your site spent 5:26 on the page.”  I often get a strange look at this point.  Then I start to explain that Time on Page is actually calculated by page jumps.  So, if someone reaches a certain page and exits from that page (or bounces), then Time on Page is 0:00.  No, that’s not a typo.  Time on Page is 0:00 whether they spent 10 seconds on the page or 10 minutes.  I’ll explain more about how time metrics are calculated below, but I wanted to give a quick example now so we’re on the same page.

time-metrics-stepping-stones.jpg

Note, although I’m focusing on Time on Page and Time on Site in this post, there are ways to calculate time metrics like Elapsed Time for certain processes that occur on a page.  For example, check out my tutorial for using TimeTracker to calculate elapsed time (or Time to Complete).  It’s an interesting way to extend event tracking to record how long it takes to finish a process like completing an elaborate form.

Access To Data Can Be Powerful But Interpreting Data Incorrectly Can Be Dangerously Powerful

I never want to rain on anyone’s parade, but it can be dangerous to walk around your office telling everyone something that might not be entirely accurate (and can be way off).  When I start helping clients with analytics, I typically make a point to grab a conference room with a white board and start explaining how Time on Page and Time on Site are calculated.  After I explain how the metrics are calculated, there are times that I hear a level of skepticism, since there isn’t hard evidence that the metrics are really calculated this way.  And I’m totally cool with that. I’d be a hypocrite to say that being skeptical is wrong, since I test everything.

In online marketing, I don’t necessarily take anyone’s word without testing it myself.  Which brings me to this post…  Based on what I explained above, I decided to run an experiment to test Time on Page and Time on Site in Google Analytics.  My goal was to document the hard numbers so you can see how the metrics are reported.  I think it’s important to understand the limitations of the metrics, so you can provide context in your analysis.  If you provide context, then your clients can better understand the analysis you provide.

The G-Squared Lab

When setting up this experiment, my goal was to isolate directories and pages and then document the exact time that I spent while visiting those pages (and sets of pages).   If each visit was isolated, then I could drill into the reporting to see exactly how Google Analytics tracked time for each page and each visit.  I’ll detail the paths I took in the experiment below.

Quick Introduction: The Formulas for Time on Page and Time on Site:

Before we hop into the experiment, let’s start with a quick introduction to how Time on Page and Time on Site are calculated.  Both metrics are based on page jumps (or moving from one page to the next).  By calculating the difference between visiting one page and the next page during the visit (via the page timestamp), Google Analytics can determine how long you stayed on the preceding page.  Calculating Time on Page via page jumps is where a major limitation is exposed.

That’s because if you visit a page and then exit the site, the Time on Page is 0:00.  As mentioned earlier, that’s whether you spent 10 seconds or 10 minutes on the page.  On a similar note, if you visit just one page on the site and leave (which is a bounce), the Time on Page is also 0:00.  Time on Site is calculated by adding the time that a visitor spent on the site by combining the Time on Page for each page that was part of the visit.  Average Time on Site simply divides total time by number of visits and does not exclude bounces.  You can see the formulas below.

Time on Page A = Page B Time – Page A Time (based on the timestamp)

Average Time on Page = Total Time on Page A/(Pageviews – Exits)

Total Time on Site (For a Specific Visit) = Page A + Page B  (or the sum of Time on Page for each page in the calculation)

Glenn Gabe
Featured SEO Writer for SEJ Glenn Gabe is a digital marketing consultant at G-Squared Interactive and focuses heavily on SEO, SEM, Social Advertising, Social Media Marketing, and Web Analytics. Glenn has over 18 years of experience and has held leadership positions both in-house and at a global interactive agency. During his career, Glenn has helped clients across a wide range of industries including consumer packaged goods (CPG), ecommerce, startups, pharmaceutical, healthcare, military, education, non-profits, online auctions, real-estate, and publishing. You can follow Glenn on Google+ here.

Comments are closed.

24 thoughts on “Tick Tock: The Limitations of Average Time on Page and Average Time on Site in Google Analytics [EXPERIMENT]

  1. @Glenn i love the post, but it also highlights another interesting aspect of a complex metric “exit rate”. It is really an annoying metric as it is also tied to how well GA is reporting bounce rates and visitors are using your site.

    Scenario 3:
    1 visitor could be tracked exiting the site after 2 pages

    Scenario 4
    1 visitor could be tracked exiting the site after 3 pages

    Exit rates seem to be useful as part of a signup process or a page that is likely to get a user interaction such as a click.

    I know i've talked about using it on projects, but its not certain if it works as it should but using time based EventTracking loading which wouldn't inflate your page views…

    1. Thanks David. I’m glad you liked my post. I found the metrics were really confusing people. I hope the results of my experiment help clear up that confusion.

      Regarding exit rate, the numbers were calculated correctly above in my experiment screenshots. It shows 100% exit rate from the pages where I exited the site. I actually wrote a post about exit rate and bounce rate, to help clarify the difference between the two metrics. Maybe this will help: http://www.hmtweb.com/blog/2007/08/bounce-rate-and-exit-rate-what-is.html

      1. Ah yes but a did have a followup test that might be interesting, it shouldn't technically affect the tracking because it works on a cookie, but how does “open in new tab” or “open in new window” impact on results?

        I ask because it is fairly common for sites to have the TARGET=”_blank” code automatically applied to links. So would these split the visitor, count as an exit or would it even possibly record it as a bounce?

  2. Thanks for an excellent post Glen. The “time on site” and “time on page” information provided by Google Analytics has confused me in the past. It is very true that decisions based on incorrect information can be damaging. Your post clarifies the issue and will greatly assist me with my analytical interpretations going forward.

    1. Thanks Heinrich. I'm glad my post could clear up some of the confusion with Time on Page and Time on Site. I think the numbers are hard to prove until you isolate visits like I did in my experiment.

      Regarding making decisions based on incorrect data, context is critically important in analytics. That's why it's essential to have a solid understanding of how these metrics are calculated.

  3. interesting post Glen, but i want to be able to take something concrete away from this.

    would you say it could almost be a best practice to always enable virtual pageviews for websites that have no real event or conversion to track, in order to get the most accurate results when it comes to reporting acutal time on page and time on stie, correct?

    1. Great question. Seeing how virtual pageviews enable you to track actual time on page does provide a compelling reason to do this. I think it depends on the page at hand. If you are extremely interested in understanding time on page for a specific piece of content, then I would recommend adding a vpv. That said, if someone simply closes the browser window or enters a new URL, then you're out of luck. This is why I try and focus on key actions versus time-based metrics. I hope that helps.

  4. Thanks Gabe — Very detailed description of time on page / site. The scenarios and screen captures really validate your points. Excellent post!

    1. Thanks Mike. I'm glad the scenarios and screenshots helped. I think isolating visits like I did in the experiment definitely helps make this topic clearer. There are still too many people relying on TOP and TOS without a full understanding of how the metrics are calculated!

  5. Thank you Glenn :-) I really appreciate the the way you explained this 'time on page' etc. issue… I “got” it. Your explanations and examples clearly outline how this works — I'm glad I ran across your article.
    Andy :-)

    1. Hey, thanks Andy. I appreciate it. I'm glad my post could help. Understanding how analytics packages calculate time-based metrics definitely helps provide context when analyzing site performance. Without context, it's easy to think that the content with the highest Time on Page is the most engaging. And that's unfortunately not the case!

  6. That was excellent. It is always a pleasure to read something new – it doesn't happen every day in this filed. Also, I really appreciate the fact I can read the whole article on Google Reader and not just a simplre preview. That's a bit OT but still, many people really enjoy that feature.

    1. Excellent, thanks Erica. I'm glad my post was helpful. I definitely plan to write more posts about web analytics in the future, so stay tuned.

    1. Thanks Dennis. I wanted to make sure that everyone had a good understanding of the metrics and the calculations before showing the results from my experiment. I'm glad you made it to the end. :)

  7. Excellent, thanks Erica. I'm glad my post was helpful. I definitely plan to write more posts about web analytics in the future, so stay tuned.

  8. Thanks Dennis. I wanted to make sure that everyone had a good understanding of the metrics and the calculations before showing the results from my experiment. I'm glad you made it to the end. :)

  9. Ah yes but a did have a followup test that might be interesting, it shouldn't technically affect the tracking because it works on a cookie, but how does “open in new tab” or “open in new window” impact on results?

    I ask because it is fairly common for sites to have the TARGET=”_blank” code automatically applied to links. So would these split the visitor, count as an exit or would it even possibly record it as a bounce?

  10. Great article, this made it very clear why the times are 0:00 when tracking our embeded flash players on other sites – there is no Page B to compare the stamp to.

  11. This is a great post. I'm in the midst of putting together a web analytics reporting dash board as I was suspicious when my analytics report was detailing 100% bounce rates but not time on that page. I thought this cant be right. This post has cleared it up form me. Nice one Glenn :-)

  12. Excellent post, really enlightening. It explains all those 0:00 on my blog (if google analytics still calculates that way that is?) it was explained in an easily understandable manner and I only installed google analytics yesterday!