Mobile Apps and Dark Traffic: Understanding where your traffic is attributed

Over the past few months, we at Avenue, have been working hard to investigate “dark traffic” in Google Analytics, and to better understand where website traffic is coming from and how it is being attributed. As a digital marketing and SEO agency, understanding the tools we use and how we consume their data and information is important to delivering the best suggestions for our clients.

Helen Pollitt, our Head of Digital, has been delivering several talks on this important topic, as we believe that the problems of incorrectly attributed traffic can be devastating to marketing campaigns.

To this extent, we have recently been researching how traffic from within apps is attributed. One of the sources of traffic we have been investigating is from the Google Quicksearch box that comes from the Google App. You might not know it by name, but it’s this search bar widget that you can put on your mobile phone home screen:

We have been experimenting with the search feature of this app to see if the traffic is coming through as direct or organic. We reasonably assumed that traffic coming through this app would be classified in Google Analytics as organic After all, it is a Google search app.

We were wrong.

We are still in early testing for this, however, we have noted that one of the phone configurations in the office seems to be sending traffic through as direct/(none) where other configurations are sending the traffic through as we assumed, as google/organic.

As can be seen from the table above, a lot of the results are as expected, however, the Samsung J3 running Android V8, Oreo, and also having a non-64 bit OS version of the Google app has traffic coming through as direct. Whilst the potential number of use cases of this are slim, there could be a number of other configurations and combinations from other mobile phones, or users not wanting to upgrade where this could become an issue.

Although this list is by no means definitive, it highlights that app instances of browsing can very easily become dark traffic that we can not track in Google Analytics, or are being misattributed to the wrong channels.

The Google Quicksearch Box experiment

We understand that clarity of data is extremely important, even more so when measuring the success of organic traffic. Knowing where your traffic comes from and how it behaves can make or break a campaign that you are running.

Performing this test, we worked under the assumption that it was either the non-64 version of the app (limited by the processor in the mobile phone, read more at APK Mirror) or the Android version itself.

From the results, we can rule out the Android version, as both the Samsung S7 Edge and the Wiley Fox phones were running older versions of Android, and equally, we can potentially rule out the app needing to be .arm64, as one of the Bluestacks instances (an Android phone emulator) was running an old app version that was just .arm.

We can also potentially rule out an old version of the app causing these tracking issues, however this is also on the Bluestacks emulator, which doesn’t necessarily mean the results are true to life.

A note on emulation

Emulators can cause problems with data, as they are not the physical mobile phones that a customer will be using, and legitimate phone data will always trump any emulated data, however, we can get a rough idea of what is occurring whilst using an emulator and how the apps interact with websites.

The takeaways for now...

Whilst this is a very specific result, we are currently working on the assumption that a combination of the Android version, as well as the .arm version of the app is causing mobile traffic to come through as direct. Whilst this may only be a very specific case, shining a light on how app, and in-app web browsers, interact with our search traffic is exceedingly important. Progressive web apps and other applications that allow browsing of the internet could be legitimate ways for users to find your content organically, and to have this attributed incorrectly will muddy data and could give you bad results on what could otherwise be a successful campaign.

From here we will continue exploring different combinations of the quickbox and Google app, in combination with different Android versions to see if there are any other phones that are sending traffic through the Google app as direct.

It is also important to note that none of the tests encompassed Apple or Windows phones, so additional testing can be done there.

We will be updating this article, or potentially creating a new one as we have more test results. If you would like to be involved in the test, feel free to test navigating to a website you can see the analytics information for, and making a note of the following:

• Android version
• Google App version
• Phone model

When you navigate to the website, let us know what your traffic comes through as at the Avenue Digital Twitter account.

Author: Chris Parker