Solution: segmenting by screen sizes

  • 21 December 2023
  • 1 reply
  • 282 views

Userlevel 4
Badge +7

At Amplitude, gaining granular insights into user behavior is pivotal, and while we may not automatically track a 'screen size' property, you can always use our default properties to build this kind of analysis. 

Our alternative properties, like ‘Platform’, ‘Device Type’, ‘Device Family’, etc can help you with understanding your user's device dynamics: 

yBXsGy8-JHtLwi7X2NFb_fam3on0t1fnNqunJkjCmewKHnYCWigjJKieOZc_ZFWVqvFpHJOM5LkaWhVVJ9f149e9TydsFMA6_Z5_2ThtZXbZEy3NrPRkkv0k2er5eggSnWzSyo6gn_66FBYNL304-Kg

You can find the definitions of each property here.

As an example, one way to analyze behavior based on screen sizes would be a segment-driven approach: creating distinct segments for each of the device types, leveraging the ‘Device Family’ parameter: 

xHTWquuScS0Xbod74cR6R8xbGOED3Un7EnQbEUvHVbv8Vy-dGQPbJ7f7tYiedAmMx4g9qiKQLFlmyahXE8nhWrry9Q0we4bJHfjKqglZcxx9fnSedamJiw98edAA-dOhAVzIV1U8_DGo4QKCZFzz0_c

Once you create your segments, you can save them and use in your charts: choose your pre-saved segments effortlessly from the dropdown list for seamless future analyses.

 

You could also instrument and capture 'screen_size' as custom properties

You can collect this data on your side (on the backend/frontend), and send it over to Amplitude as event or user properties (e.g ‘screen_width’ and ‘screen_height’). 

For example: 

If ‘screen_width’ collected = 1500 & ‘screen_height’ collected = 800, you can now set filters for events with conditions based on these measurements (like screen_width >/=/< 1500 & screen_height >/=/< 800 etc).

Hope this helps with analysing your users’ behaviour based on the screen sizes they use! 

 


1 reply

Hi

Thank you for suggesting this approach, but I must admit I'm rather perplexed by it.

The reality is that relying solely on default properties falls short when it comes to effectively segmenting by screen sizes or distinguishing between web desktop and mobile users. The issue lies in the sheer volume of values for device family, device type, carrier, or operating system. With hundreds, if not thousands, of values, selecting the relevant ones for each segment becomes an impossible task.

While the Platform property proves to be useful, the granularity of the other properties renders them impractical.

For instance, when attempting to create a segment for web desktop users, we can narrow it down to device families like Windows, Mac, and Linux. However, there are numerous other device family values, such as Google Chromebooks or miscellaneous categories, making it challenging to determine whether they classify as mobile or desktop devices when presented in a lengthy list.

To illustrate further, if I were to apply this solution to all our web traffic, I would only be able to classify approximately 70% of it as either desktop or mobile (inclusive of tablets). The remainder of the traffic falls outside this classification, rendering it totally unusable for any meaningful analysis.

Nobody working with or at Amplitude has been able to figure out a better way yet?

Reply