CJA: Unexpected Events Returned When Metric Is Filtered by a Dimension | Community
Skip to main content
Level 2
July 24, 2026
Question

CJA: Unexpected Events Returned When Metric Is Filtered by a Dimension

  • July 24, 2026
  • 12 replies
  • 139 views

I'm seeing an unexpected behavior in Customer Journey Analytics and would appreciate some clarification.

When I use custom event dimension along with the Events metric, which is filtered with specific attribute (for example, where a specific attribute equals a particular value), the report still returns events that don't appear to be associated with that filtered attribute.

I verified both the incoming analytics and CJA request and the data layer, and I couldn't find the filtered attribute being set for those events. Based on that, I would expect those events to be excluded from the results.

Can someone explain why these events are still appearing when the metric is filtered by that dimension value? Is this related to how CJA applies filters at the event, session, or person level, or is there another aspect of CJA processing that could cause this behavior?

12 replies

JasonEg2
Level 2
July 24, 2026

Is it possible that your dimension/component in CJA has a setting for persistence/attribution in the data view that would cause the events to be attributed to other dimension values? 

 

Jason Egan, Principal Solution Architect @ Adswerve
Level 2
July 29, 2026

Thanks ​@JasonEg2 for your response. I have not set the persistence for the dimensions 

Level 5
July 28, 2026

The behavior you're seeing is usually related to the scope at which the filter is evaluated in Customer Journey Analytics.

In CJA, filters (segments) can be applied at three different levels:

  • Event – Only the events that meet the filter criteria are included.
  • Session – If any event in the session matches the filter, other events from the same session can also be included.
  • Person – If the person qualifies for the filter, events associated with that person may also be included.

If your expectation is to count only the events where the attribute has a specific value, make sure the filter is evaluated at the Event level. If the filter is using a Session or Person container, it is expected that additional events from the same session or person may appear even if those individual events do not contain the attribute value.

A few things to check:

  • Verify that the filter uses an Event container.
  • Confirm that the attribute is present on the same event that you're counting.

If you've already confirmed that:

  • the attribute is not present on those events in the source data,
  • the CJA request also doesn't contain the attribute,
  • and the filter is scoped to Event,

then this would not be expected behaviour. In that case, it would be worth reviewing the Connection/Data View configuration in more detail or opening a support case with Adobe so they can investigate the dataset and query.

According to Adobe Experience League, the scope of the filter (Event, Session, or Person) determines which data is eligible for inclusion in the results, so verifying the filter container is the first step in troubleshooting this type of issue.

Level 2
July 29, 2026

Thanks ​@Pankaj_Sabharwal  for your response. 

I made a metric from dimension and not by creating segments. 

I found that one of the launch rules included a "Clear Variable" action, but the Clear checkbox within that action was not selected. Will this cause the value to persist?

Level 5
July 30, 2026

Yes, that could be the cause. If the Clear Variables action is configured but the Clear option isn't enabled, variables may persist across subsequent Adobe Analytics tracking calls if they aren't updated with a new value. This can result in a previous dimension value being sent on later hits, even when it's no longer present in the data layer.

To verify this, inspect the network requests using Adobe Experience Platform Debugger. If the value persists across hits, enable the Clear Variables action or reset the variable after the tracking call

Jennifer_Dungan
Adobe Champion and Community Advisor
Adobe Champion and Community Advisor
July 28, 2026

Agreed, while I don’t use CJA, my first thoughts were what persistence / allocation the dimension had (like ​@JasonEg2  mentioned).

 

@TapaswiniPa1 I assume you “made a metric from dimension”? And you are not creating segments at this time?

Level 2
July 29, 2026

@Jennifer_Dungan  Thanks for your response. I made a metric from dimension and not by creating segments. 

I found that one of the launch rules included a "Clear Variable" action, but the Clear checkbox within that action was not selected. Will this cause the value to persist?

 

 

Jennifer_Dungan
Adobe Champion and Community Advisor
Adobe Champion and Community Advisor
July 29, 2026

Hi ​@TapaswiniPa1 this is hard to tell without seeing the implementation.

 

I know that isn’t a great answer on its own, which is why I am now going to dig a little into what that means….

 

The “Clear Existing Value” in this context is really just a part of how websites work. If you set a global object to a value of “X” then that object will be “X” until it’s overwritten (either with a new value “Y” or cleared to be empty “”) or the page is reloaded.

 

What this means for Adobe depends on if your site is a traditional design, or a Single Page Application; as well as if you have a lot of clicks/actions that take place on your pages.

 

To help explain this, let’s look at a simple flow:

 

  • Page A Loaded
    • value set to “A”
  • Page B Loaded
    • value set to “B”
  • Click on Page B
    • value is still “B”
  • Page C Loaded
    • (no value set)

 

If the above is a “traditional site” with every page loading fresh, then the “clear” really isn’t doing much, the data won’t persist into the next page (there are caveats to this). On the click on Page B, the value will still be “B” and IF you include the value in your tracking, then “B” will technically be sent on the tracking call (you will also see value “B” when you are debugging - this will be treated like a new instance of the value). There will be no value on Page C (as the natural behaviour of the web will wipe the values anyway).

 

If you are using an SPA, and the page is not reloaded… by not setting a value on Page C, the previous value of “B” will send again, as it’s still set in the webpage. You will see “B” in your debugging as a new instance.

 

 

Now, of course, launch has additional “persistence” on the Data Element level.

You can choose “None”, “Pageview”, “Session”, “Visitor”… if you have set a Data Element to a value, and chosen a “Session” duration, then Launch will keep the value until it’s overwritten for the rest of the session, and this value can be sent on your tracking calls on every subsequent hit (clearing the values as you showed should have no impact on this… as the “Clear Values" is about clearing the tracking Data Layer not the Data Elements )

 

 

Then of course, coming back to the “Dimension Persistence” of the original discussion… this is when a value is sent to Adobe, and despite not seeing this value being re-sent during debugging, Adobe persists the value based on the settings… if it’s Hit, the value will only exist where it was actually sent, if it’s Visit, the value will be empty until the first hit in the visit (the instance) and then the value will exist on every hit for the rest of the visit (or until a new value is set, then that new value will persist).

 

 

Since in your original post, you seemed to indicate that you are not seeing the values during debugging, its seems unlikely that the issue lies with either your Data Element duration or the lack of Clear Variables… as both of those cases, you would still see the value being sent in your tracking.

Level 2
July 31, 2026

Hey ​@Jennifer_Dungan  Thanks again for your response. I have checked the data element and could see that the storage duration is set to “None”

Josh Stephens
Community Advisor
Community Advisor
July 30, 2026

Have you broken out the metric by the dimension that was used to create it? I’d be interested to see what the breakout shows.

Does the metric have any any custom attribution or other settings beyond include/exclude?