Analytics - Interpreting and understanding metrics

Let's start with a fact.

"On Google Analytics, the Average Session Duration is not your website's Average Session Duration."

It’s easy for us to log into analytics platforms such as Google Analytics, see a number and assume that it means what we think. In reality, there are often nuances to the definition of each figure.

I was in a meeting in the last month and we were discussing the bounce rate of media activity. Someone asked “How long did the people who bounced spend on site?”. The answer, when using Google Analytics and no additional wizardry, is “We don’t know”. I realise that a number of people will already be aware of this, but think it's good to highlight these types of considerations.

“Hey Sam, how long did people spend on our website?”

If you’re asked this by a colleague, you may just log into Google Analytics. On the main Audience screen for the analytics account, you'll see the following: -

So, Google Analytics says that the average session duration is 2 minutes and 28 seconds. However, the truth is a little different. 

An example

The below example is a simple journey on a website, from a landing to an exit.
The time of each interaction is at the bottom of each step.

There are two questions we can ask from this example: -

1. How long did a user stay on your website? 
   4 minutes
2. How long will Google Analytics say the user spent on your website?
   2 minutes

Why is this? Because the only way that Google Analytics tracks time is by stamping the time that the user views each page. 

• At 10:00, the first pageview is recorded by Google Analytics
• At 10:02, the second pageview is recorded by Google Analytics
• At 10:04, the user exits the site and goes somewhere else.
  This isn't recorded by Google Analytics.

On an exit, there is no time-stamp. Google Analytics doesn't know that the user is no longer on the page, so the time spent on page two isn't recorded.

Bounced Traffic

Whichever way we view average session duration, we know that the figure on Google Analytics is pessimistic, and under-represents the real number.

By including bounced traffic (sessions with only one pageview) for instance, the numbers are massively skewed.  To counteract this, we can segment the traffic into two segments: -

• Segment 1: Bounced Sessions
  Sessions with 1 pageview
• Segment 2: Engaged Sessions
  Sessions with more than one pageview

The below image is from the same Audience report as earlier, but with the above segmentation in place.

This shows the average session duration of bounced sessions is 0 seconds, whilst the average session duration of engaged sessions is over 4 minutes.

So, originally we may have told the colleague that people spend 2 minutes 28 seconds on site. Once we remove the bounced sessions the actual average session duration of sessions is over 4 minutes - over 70% longer.
This is still a pessimistic stat. The time spent on these users' last pages aren't included, but the skew is reduced, and the figure is more realistic

In summary

Time-based metrics on out-of-the-box Google Analytics are flawed. As long as you are aware of these flaws, then they can still have value during analysis. There are arguably much better ways to measure 'success' on non-transactional sites. There are also workarounds, which are only a Google search away, but will require custom analytics set-up and a developer to implement. 

Each metric on Google Analytics (and other systems which report results) has a definition. It's important to understand what the numbers mean and what they don't mean, especially if you're relying on these figures in order to make business or strategic decisions.

Need guidance with your analysis?
Contact RBH's James Brazier jamesb@rbh.co.uk for a chat.