Skip to main content

Hi!

I found that Event Property filter doesn’t work properly and doesn't show me the actual info. In my case it’s the User's 'initial_referring_domain' Property but it seems to be the general problem.

What I see (screenshots are below):

  1. All Events for November’2022 → 235 Users
  2. Events for November’2022 with ‘empty’ value of the Property →  235 Users (that means that all values are ‘empty’, right?)
  3. Events for November’2022 with ‘hypetrain.io’ value of the Property → 0 Users (looks consistently, yeah)
  4. Export (‘Download Users’) for the above case #1 (All Events for November’2022 w/o filters) → tons of Users have ‘hypetrain.io’ value, the most of values for this column aren't empty!

I checked this for October as well - at least 14 Users have the ‘hypetrain.io’ value in exported .csv-file not 7 ones.

I’ve tried to use Filter by Event Property ('initial_referring_domain') in both sections (‘Events’, ‘...performed by Users’) of the Chart but the results are the same.

All Events, Nov’22

 

Filtered by Property (‘empty’) Events, Nov’22

 

Filtered by Property (‘hypetrain’) Events, Nov’22

 

All Events, Nov’22 - exported

 

 

How I can fix it?

Thx for your attention

Hi Sergey, 

Thanks for your post and for the screenshots! I think what is happening is that the event is not being sent with the value in November. The unique users firing the event are falling in the (none) bucket, which means at the time of the event there was not a value. 

If the value is sent with a later event, it will appear on a .CSV download under the category. But in this case, it looks like the value isn’t received by Amplitude when the left side event is fired in November. 

Here’s a help center article with more details about the (none) bucket. 

Regarding a fix, you can’t modify events that have been ingested, but there is this potential workaround found in the FAQ. 

Let me know if you have questions. Thanks! 

Best, 
Loc


Thank you for answer, @Loc Tran 

But I’m disagree with the following:

... I think what is happening is that the event is not being sent with the value in November. The unique users firing the event are falling in the (none) bucket, which means at the time of the event there was not a value.

If the value is sent with a later event, it will appear on a .CSV download under the category.

That’s why:

  1. In my case it looks like the value of the ‘initial_referring_domain’ property for this ‘user-created’ event stopped (!) to appear AFTER October (from chart view perspective, see chart screenshot below) ...
  2. … but from the ‘Download User’ .CSV perspective it shows that value appeared since September till December (regarding my ‘new user - this specific registration event - it’s date’ data in my own DB here I know month of user pop-up, read below), but it looks the same from the ‘Export CSV’ .CSV perspective (yes, it coincides with the above-mentioned chart, see screenshot)
  3. The type of this event (user registration) define the uniqueness of every user not cross ‘buckets’ but within the whole bunch of events (only one event for one user forever), and the value of property ‘initial_referring_domain’ couldn’t be added LATER from this type of event (‘user-created’) if it can be added from this type of event only
  4. It’s the Amplitude ingrained user property so I can’t even influence the first/next/last/any firing of this event with non-null property value - it’s the Amplitude mighty and selling point to gather and deliver this info without my own manually-developed and tailored ‘stalking’ code

It can't be the ‘NO-BUG-BUT-FEATURE’ situation, because in the User CSV-export file I have much more unique values than on the charts and all new events have ‘(none)’ values.

Or do I not understand how the ‘initial_referring_domain’ property should be attributed to my Users by Amplitude?

No any non-null values except one month (October), charts

 

No any non-null values except one month (October), CSV

 

 


Thanks for following up @Sergey! I wanted to confirm, were you able to make progress and achieve what you set our to to do? Please keep us posted! 😁

 


Hi @Sergey

Thanks for your reply and the screenshots. Are you still seeing a discrepancy in the ‘hypetrain.io’ value? 

If so, let me know and I can private message you for the Org ID and an example chart so I can look into it. We don’t want to share any identifying information here in the thread! 

Thanks,

Loc 


Reply