Event Segmentation, group by user property vs event property

  • 3 February 2022
  • 4 replies




I'm trying to group by a event by the current value of a user property value.

It feels like the event is only grouped by the value of this property for the event but not the user property value.

I would like the breakdown available in User composition but in Event Segmentation.

How can I achieve that ?








Best answer by Dr.Data 3 February 2022, 22:21

View original

4 replies

Userlevel 2

@tthias could be wrong here, but based on my experience the event segmentation chart doesn’t allow for you to define if the user property value you are grouping by is the most recent, cross project, or anytime value like the user composition chart.  Depending on your range of user property values you want to group by, you could create cohorts for each of the values or range of values you are trying to monitor with the most recently clause in the cohort definition and then input those cohorts into an event segmentation chart.  



Userlevel 4
Badge +7

Hi @tthias

Like what @Dr.Data has mentioned, in the event segmentation chart (and most charts in Amplitude), the user property value queried will be on the event-level, and not based on the most recent value. Are you on our Scholarship or paid plans? If so, the suggestion of creating cohorts to segment these users based on their most recent value for the property will help! If not, maybe you can elaborate a bit on your objective of creating the charts, and we can help brainstorm if there is a workaround. 


Hey, thanks for the replies !

So my use case is the following :

Anonymous users land on our website with an utm_source stored has user property.

Then creates an account, and the account creation event is server-side.

I want to break down the account creation by utm_source.


Do you think that’s doable ?

Userlevel 7
Badge +10

Hey @tthias 
Since your utm_source is set as a user property, the easiest way to break down that property would be to use a User composition chart similar to this.