Google Analytics: Misunderstood Metrics27th August 2017Analytics
There are a lot of examples of Google Analytics reports which begin by looking at common KPI metrics and draw conclusions such as the following:
February has been a bad month for our website. The overall bounce rate has increased by 15%, average session duration has decreased by 30 seconds and there are 10% fewer returning visitors.
This seems perfectly reasonable, but does this portray an accurate picture of what's really happening? To figure this out, we need to take a lot of some of these metrics in a little more detail.
Bounce rate is the percentage of single page visits. If someone visits your site but leaves before visiting another page then they are said to have bounced. A high bounce rate can often suggest that people are not engaging with your site and that something is wrong.
However, this isn't always the case. Blogging sites often have high bounce rates as people often read one specific article before moving on to somewhere else. A user could be spending a good 10-15 minutes enjoying a news article but then will leave the site which will be recorded as a bounce.
When a news or blog post you've written gains traction (for example on social media) you could end up with hundreds of new visitors reading that one post. This will cause a huge increase in your site's bounce rate, but in this case it's clearly a positive result.
Similarly an increase in bounce rate after a site design could be an indication that the content has a better structure. Users can quickly find what they need without having to spend time searching through page after page. For sites providing support or help articles, a high bounce rate is seen as a positive outcome.
Average Session Duration
Average session duration in Google Analytics is calculated by taking the total duration of all sessions and diving this by number of sessions. So for example, if the total session duration in a day was 10,000 seconds and a site had 100 visitors then the average duration would be 100 seconds (reported as 00:01:30).
Unfortunately, this is a bit misleading as the way GA determines the duration of a session is a little different to what you might expect. Many people assume that once a user visits your site there's a kind of ticking clock in the background that is recording the duration of the session when actually it's a lot cruder than that.
When a user visits a website, a hit is sent to Google Analytics. They then visit a second page and another hit is sent. The session duration is the time between these hits. There is no hit sent when someone leaves a site, so the last page they visit isn't counted in the calculation.
To illustrate this, lets say a user visits 2 pages. They spend 20 seconds on the first page, 4 minutes on the second page and then leave the site altogether. There is a hit sent for the first page (at 0 seconds) and a hit sent for the second page (at 20 seconds). So the total duration is calculated as 20 seconds even though it was really 4 minutes and 20 seconds long!
Similarly, any time a user visits just one page of your site, the only hit sent is at 0 seconds. So regardless of how long they were on that page, GA will report the session duration as 0 seconds (note this may be different if you have events set up on that page).
The average session duration metric uses all of these results when it makes its calculation, Which means it is often reports a time which is quite far from the truth. If your site has a lot of bounced visits then they'll be an even bigger discrepancy between these values.
New Sessions and New Visitors
A session in GA is a group of user interactions within your website that take place within a given time frame. This means if someone visited a bunch of pages, downloaded some documents, completed a sign-up form this would all be counted as a session.
However, if a user is inactive for 30 minutes at any point (this value can be changed) then a new session is started. For example, if someone visits an e-commerce site decides to buy a product but waits until they've eaten lunch before purchasing, then these will appear as two separate sessions. A new session is also created for anyone browsing a site when the clock strikes midnight or if a user leaves the site and returns using a different entry method (e.g. advertising rather than search).
GA is able to keep track of these sessions by setting a series of cookies in a user's browser, which last for a long time. The problem with cookies is that they can be deleted. If you have ever cleared your browsers cache and ticked the box which says 'delete cookies' then you will now appear as a new visitor for every site until a new cookie is set.
Also cookies are linked to the browser or device you are using. So if someone visits a site on their phone, then completes a purchase on the same site using their desktop computer then these are counted as separate users.
Another problem with cookies is that it's becoming more common for users to install ad blockers. Although most people use them to help reduce the number of intrusive pop-ups and online advertising many of them also block GA cookies by default, which means that these users do not show up on any analytics report. Once study suggests that this number may be as high as 11%.
Is it all doom and gloom?
We could conclude from all this, that we shouldn't use data from GA as it doesn't always present an accurate picture, but this certainly isn't the case. All methods of data gathering have inherent flaws, but being aware of the problems allows you to take account of their effects and still present meaningful and useful data.
By having a greater understanding of how metrics are calculated we can draw much more accurate conclusions and make better predictions for the future.