Analysis of “Average Session Duration” in Google Analytics

There are many reports in Google Analytics that include “average” metrics: average of pages/session, average bounce rate, average session duration, average time on page, etc. etc. Average values can be very useful in analysing patterns and trends in our data, but we also need to be mindful of some of the risks and shortcomings inherent in these metrics.

This blog post will help you to understand and use the “average session duration” metric in Google Analytics.

Can we trust this data?

The simple answer is that we cannot trust this data – why? –  because it is an average of metrics which are themselves averages of other metrics which are themselves averages of other metrics, and so it goes on. For example:

Google Analytics Channels report

Here (in June 2014) we see that the site-wide average session duration was 1 minute and 16 seconds.  The average session duration for traffic coming in from Google organic search is 53 seconds.  The average session duration for direct traffic is 1 minute and 50 seconds.  The average visit duration for traffic referred from nl.visma.com is 3 minutes and 43 seconds, and so on.  Already we begin to see that the site-wide average session duration is not as straightforward as it might appear at first glance.  In addition, as with all average metrics, any outlying (extreme) values can dramatically skew the results.

So how should we handle these metrics, and can we gain any value from them?   Rather than relying on the site-wide average session duration, we analyse and report the “average session duration” for each individual traffic source.  This gives us a far better insight into the true “average session duration”, but it is still not the most efficient way to deal with averages.

In order to better understand the  “average session duration” for the traffic, for example, coming in from  “Google organic search”, we need to further segment this traffic source:

average session duration
This is the “Channels” report which you can find in Google Analytics under the “Acquisition” drop-down in the left side navigation menu.  In this example I have already clicked “Organic Search”  and also chosen “Source” as the Primary Dimension (the default is “Keyword”).  The report displays all the traffic sources (search engines) that are sending organic search traffic to our websites.  The majority (93%) of our organic search traffic is coming in via Google, so it would  be useful to further explore the “average session duration” for this traffic source.

Staying with the “Channels” report, select “city” as a secondary dimension:

average session duration

We can now see the average session duration for Google organic search traffic, for each of the cities where these visitors reside.   Let’s suppose we are planning a city-specific campaign (e.g. recruitment). and we want to understand the on-site behaviour of visitors from Trondheim before targeting this location with PPC and other marketing activities.  The “average session duration” for these visitors will provide a valuable insight into how we should plan and execute our campaign(s).

The next step then is to create a custom advanced segment (a topic that I will cover in a later blog post) which reports only the Google organic search traffic from Trondheim, and apply this to the “Session Duration” report (found under Audience > Behavior > Engagement):

average session duration

In this example we can see that 53% of the sessions last for 0-10 seconds, and approx. 24% of the sessions last for 61 or more seconds.  If we assume that it would take at least 3 minutes for a visitor to convert on our website, then only about 13% of the Google organic traffic from Trondheim has the potential to convert.  This type of distribution analysis gives us a far more accurate insight into traffic coming from Trondheim than the average session duration of 33 seconds that we saw in the previous report.

There is another issue worth mentioning which can also skew the “average session duration”, and that is data sampling.  Google Analytics uses data sampling to analyse large data sets in an efficient and timely manner.  If your Google Analytics reports are affected by data sampling then metrics such as “average session duration” can have inaccuracies of between 20 and 80%.  It is important in these cases to maximise the sample size (usually by choosing a shorter date-range) before using any average based metrics for analysis.  I will discuss ways to improve accuracy in Google Analytics reports in a later blog post.

In conclusion, the decisions we make and budget allocated are (or should be)  influenced by the analysis, interpretation and reporting of user behaviour on our websites.  For this reason (amongst others) we need to be careful and thorough when utilising average based metrics such as the “average visit duration”.

About

Originally from Liverpool, UK, Ilan was a pioneer in the field of website analysis and brings to Visma more than 10 years of digital marketing experience. Today he is responsible for the tracking, measurement and analysis of web traffic data across the entire Visma ecosystem. Outside of the work environment, Ilan can often be seen driving (very slowly) around Oslo in his 1961 VW Beetle.
Connect with Ilan: