Solved

Bucketing users by properties for funnel analysis / How do User Filters work


Badge

Hi everyone, I’m new here and this might be a well known answer – I tried searching but wasn’t able to find one.

 

My high level goal is to set some user Amplitude properties that I can later on use to bucket users into categories and use that for funnel analysis. The problem is at user signup I don’t have the values for those properties yet (for example the user hasn’t shared their location yet)

 

I wanted to understand first what those two user filtering properties mean:

 

Is the following correct for the following image: 

Any Events that are performed by any user who at the current time (not the time of the event) their User Property Country = “United States” (i.e. a user who had an empty Country at the time of the event but their current User Property Country = “United States” will be included)

 

Filter applied to event #1

 

 

Is the following correct for the following image: 

Any Events that are performed by any user who at the current time (not the time of the event) their User Property Country = “United States” (i.e. a user who had an empty Country at the time of the event but their current User Property Country = “United States” will be included)

 

Filter Applied to both events

 

 

That means that for funnel analysis both of those charts are identical? 


 

 

icon

Best answer by Saish Redkar 16 May 2022, 07:24

View original

4 replies

Userlevel 7
Badge +10

Hey @guenena 

Welcome to the Amplitude Community!

I’m not sure if I completely understood the funnel you are trying to set up here with “Any Event”, but I’ll try to answer your core question of user properties first.

Since, Amplitude is an event based analytics platform,  all charts and cohorts query at the event level. Charts will return the user property value at the time of an event. You can read more on that here and here.

So your User Property Country = “United States” will be applicable at the time of the event.

In your chart #1, the Country property filter will be applied to only event 1 in your funnel.

In your chart #2, the Country property filter will be applied to both event 1 and event 2 in your funnel.

If you want to focus on users whose current User Property Country = “United States” , then one of the ways would be to create a cohort ( if you have access to this feature ) and use that in your chart.

The most recently clause will select only users whose most recent value for Country user property is United States. This will allow you to include users who didn’t have the desired property during the initial few events, but had them on their recent events.

 

Let me know if I have interpreted your question correctly. Hope this helps.

Badge

You interpreted the question correctly. Thank you, that is not what I thought which means I have been incorrectly reading Amplitude charts for 2+ years. Thanks a lot. 

 

The Any Event Funnel, I basically wanted to check how many users who have taken a certain event but I don’t know what event to start the funnel with. For example check how many users completed their profile bucketed by country, so I want to latch on any event that has the country information and it’s at the top of the user funnel.

 

However, with your answer I can see I can use Cohorts to bucket the users and then check how many users of that cohort have completed their profile. Let me know if there’s another way of achieving that (?)

Userlevel 7
Badge +10

Ideally, a “Completed Profile” kinda event will help in understanding this funnel a lot better rather than juggling with Any Event, and then using the Cohort segmentation.

If not, do you have any particular user property which gets set once a user completes a profile?

Not the most accurate way, but an example funnel for your new users could be (Step1) Any event where this _propertyA = (none) → (Step2) Any event where this _propertyA is not (none) or the actual value.

Badge

I see got it, we have been using events, but adding them as user properties makes a lot of sense. Thank you Saish. I did ask that question exactly in another thread and you answered it here. Thank you for going above and beyond @Saish Redkar :) 

 

 

Reply