May 8, 2015

What are Devices in Google Analytics?

In a recent Google Analytics post, we looked at Segmentation, and in this post, we’re going to review a similar concept - the insights we can gain by viewing Devices data.

At the highest level, Devices are broken down into:

  1. Desktop computers
  2. Tablets (iPads for example).
  3. Mobile (Generally considered phones).
     

The assumption we’ll make here is that your site has been designed for responsive and we’re comparing data by Device, but all on the same domain.

Under the Audience tab in Google Analytics, there is a Mobile > Overview option. This reveals the Device Category showing the three devices mentioned earlier.

What should you look out for?

The key things to understand is how users are interacting with your site at a device level.

  1. Sessions tells us how many visitors by device.
  2. Bounce rate reveals users that went no further than the page they landed on.
  3. Pages per session demonstrates how many pages the average user viewed during their session.
  4. Duration talks to the average time spent.
     

By observing these metrics at a basic device level, we can already tell that our users are behaving in different ways depending on the appliance used to interact with our site. Once the data has been reviewed, assumptions can be formed and tested.

What are examples of conclusions I could make?

In maintaining the theme with previous posts, we’ll analyse interactions by device applied to a blog post. Hayley posted a piece about Google’s Mobilegeddon, aptly titled: If you're not mobile-friendly, Google's not your friend and we'll use that as our reference.

  1. Desktop has 79% of all views and an average time on page of 3:30 mins.
  2. Mobile has 17% of all views and an average time on page of 10:40 mins.
  3. Tablet has 4% of all views and an average time on page of 1:41 mins.
     

A quick overview of this information suggests that mobile views are the most potent - users spending over 10 minutes on average reading. The asterisk here is that there is a lot more data available for the desktop user than the other devices. As your pageview numbers increase, your confidence in the accuracy of the data should amplify. 400 is a standard sample size that should ensure you have sufficient data to make decisions.

What should I do with this information?

The best thing to do is try a set of simple experiments. We’re looking at analytics from a blog post, so there are a few simple baselines we can establish.

  1. Read the blog and measure the time it takes to complete. Hayley’s Mobilegeddon blog took me 2:10 to read. Even if readers were slower and more considered, it shouldn’t take more than 3 minutes. The caveat here is if links that reference related stories have been included, this may explain why your blog was viewed for longer on average - as is the case with our mobile example.
  2. The average reader is taking 3:30 to read it on a desktop, so we can be satisfied with that.
  3. Tablet readers at only 1:41 are either really fast, or not as engaged as they may be on other devices. This device is also most prone to skew, representing the least robust number of views.
     

If curiosity is getting the better of you, you’re always able to go one step deeper and look at the the Hour of Day dimension. A quick eyeball demonstrates that 40% of all views occurred during 9 and 10AM. This helps to explain the desktop views as our intended readers are at work during that time, but it also challenges the theory that mobile users spent ten whole minutes on this page. In peak commute times, that would seem reasonable. After hours, also logical, but during business hours, somewhat unexpected. We can still conclude however that users who read this piece on a phone were more interested than desktop users and read the referenced articles also. We can also speculate that not all users are tied to a desk.

Tablet users may not be engaged, the content may not be practical for the level of interaction the post may demand and users reading at this time of day on their tablets may be looking for something more efficient to explore. With limited data, we still have questions as to how robust the sample population is. One approach would be to split test the send - 50% of the database receives the email at 9AM and 50% at 8PM when tablet use is greater.

Suggested actions.

  1. Try marketing posts at different times of day. As discussed in the Google Analytics segmentation post, different social media channels will also provide different user results. In addition to split testing through email sends, try different social posts at different times of day.
  2. Be sure that each device displays the content with best practice in mind. If every optimised website renders the tablet font size at 16pt and your site displays it at 8, users may find this frustrating and leave before getting sufficient value from your content.
  3. Don’t continue do the same thing expecting a different result.
     

About the blogger.
Jason Healey is bwired’s Head of Delivery. A passionate digital strategist, Jason loves mining data in Google Analytics and promoting approaches to problem solving aimed at improving your digital ecosystem. Read other posts by Jason about Strategy, Google Analytics and Digital Marketing.

Leave a comment