Segmentation not Filtering out campaigns correctly

[AlexanderRo4]

I have 3 campaign buckets that I'd like to create a segment for. The campaigns typically look like arship20240712, arorder20240711, arcataff20240712 for example. I have my segmentation setup as follows:

bucket 1

Include - Visitor > campaign 'starts with' arcataff

campaign 'starts with' arjcataff

etc

bucket 2

Include - Visitor > campaign 'starts with' arship

campaign 'starts with' arorder

etc

What happens is several of the other campaigns (arship etc) still appear in buckets that its not supposed to IE: seeing arship campaigns in bucket 1 when its only supposed to show arcataff.

Why does it continue show other campaigns?

[Jennifer_Dungan · Accepted solution]

By using a VISITOR scoped segment, you are pulling back all hits from Visitors who interacted with the campaign at any time during your reporting period.

Let's look at a simple example (I'm not sure what your expiry is on your campaigns, I'm going to assume this is tracking code and 7 day expiry):

Visit 1 (campaign X) - June 1

• Page 1 (campaign X is set)
• Page 2 (campaign X is maintained, but not set)
• Page 3 (campaign X is maintained, but not set)

Visit 2 (no campaign) - June 3

• Page 4 (campaign X is maintained from previous visit, since it's within 7 days)
• Page 5 (campaign X is maintained, but not set)
• Page 6 (campaign X is maintained, but not set)

Visit 3 (campaign Y) - June 5

• Page 7 (campaign Y is set, overriding the previous X, despite within 7 days)
• Page 8 (campaign Y is maintained, but not set)
• Page 9 (campaign Y is maintained, but not set)

Visit 4 (no campaign) - June 10

• Page 10 (campaign Y is maintained from previous visit, since it's within 7 days)
• Page 11 (campaign Y is maintained, but not set)

Visit 5 (no campaign) - June 24

• Page 12 (no campaign - nothing set and outside of 7 day expiry from when Y was set)
• Page 13 (no campaign)

Now, let's say you are pulling a report for all of June, if you are looking for "VISITOR with Campaign Y", you will get all the above pages (1-13), since your Visitor saw campaign Y on visits 3 (explicitly set) and 4 (maintained due to 7 day attribution). Don't forget, even the same visit could have multiple campaigns... I started from social media campaign, got an email while I was on the site and clicked through from that email.... 2 campaigns in 1 visit.

So what are you trying to find? Don't worry about the metric you want right now... the metric used with your segment is unrelated... Segment scope is about what hits you want to return in your data set.

Are you looking for:

• Visits that have Campaign Y attribution?
• Or Visits that explicitly came via Campaign Y
• Just the entry hit with Campaign Y?

Understanding your needs will help us to help you build the correct segment.

[AlexanderRo4]

That's a great explanation. Thank you.

What I'm looking for are all Visits for say "Last Week" that explicitly came via campaign Y, Z, X but should not see anything from A, B, C.

Added note: The attribution model I use is 14 day last touch.

[Jennifer_Dungan]

OK, and what happens if you have a visit with both?

• Page 1 (Campaign X set)
• Page 2 (Campaign X carried forward)
• [User gets an email with Campaign Y]
• Page 3 (Campaign Y set)
• Page 4 (Campaign Y carried forward)
• [User foes a search, gets a sponsored link with Campaign Z]
• Page 5 (Campaign X set)

This visit has explicit X, Y and Z set in the visit... should this visit be included because it has X? Or should it be excluded because it has Y / Z?

[AlexanderRo4]

I only want to see X activity, I don't want to see the Y and Z activity but I want all X visits regardless.

[Jennifer_Dungan]

So you want the "HIT" of the entry into your site, using "X"?

You can still pair this with the Visits metrics, but you don't need all the pages in the Visit?

Or you want all Hits that are attributed to "X" regardless of whether it was an entry?

So in:

Visit 1

• Page 1 (Campaign X set)
• Page 2 (Campaign X attributed)
• Page 3 (Campaign X attributed)

Visit 2

• Page 4 (Campaign X attributed)
• Page 5 (Campaign X attributed)

Visit 3

• Page 6 (No Campaign)

Visit 4

• Page 7 (Campaign X set)
• Page 8 (Campaign X attributed)
• Page 9 (Campaign Y set)
• Page 10 (Campaign Y attributed)

Entry into Site Using X

- would return Pages 1 and Page 7

- 2 PVs, 2 Visits

Pages Attributed to X

- would return Pages 1, 2, 3, 4, 5, 7 and 8

- 7 PVs, 3 Visits

(I think this is what you are looking for... all "X" interactions, and the ability to see the Visit count where X came into play)

Or something more complex like:

Pages Attributed to X ONLY where there was an entry to the site in the Visit with X

- would return Pages 1, 2, 3, 7 and 8

- 5 PVs, 2 Visits

In the above example, there are 4 visits in total, 10 pages in total, 3 pages where a campaign was set, 6 pages where a campaign was attributed, and 1 page where there was no campaign.

[AlexanderRo4]

The campaigns are representative of email campaigns. So I'm not sure the example for Visit 4 makes sense since if they came to the site through email campaign Y then that should be a different visit as campaign y set and attributed.

Visit 4

• Page 7 (Campaign X set)
• Page 8 (Campaign X attributed)

Visit 5

• Page 9 (Campaign Y set)
• Page 10 (Campaign Y attributed) 

Yes this is the outcome I'm looking for.

Pages Attributed to X ONLY where there was an entry to the site in the Visit with X 

[Jennifer_Dungan]

I wouldn't discount someone clicking through multiple emails in a session.. emails sit in people's inboxes for a long time... or on multiple ads within the same email in a session... (I don't know your set up)

So you may still see some non-matching campaigns, if that happens...

Now, because you have 14 days attribution, that will run between visits:

Visit 1 (from Campaign)

• Page 1 (Campaign X set)
• Page 2 (Campaign X attributed)

Visit2 (direct)

• Page 3 (Campaign X attributed)
• Page 4 (Campaign X attributed)

So the simple solution, where "X" is available based on your attribution

HIT [

    campaign starts with arcataff

]

This will pull back all the hits where "X" was set or attributed

But if you need to restrict X to visits where it was set:

HIT [

    campaign starts with arcataff

    AND

    VISIT [

        HIT [

            campaign starts with arcataff

            AND

            campaign instance exists

        ]

    ]

]

Basically, I am looking for all hits with "X" campaign, but I am restricting those hits to Visits, that have a hit for campaign "X" AND have the instance of Campaign metric (this is triggered when the eVar is explicitly set)

[AlexanderRo4]

Okay then that's the difference I didn't understand between GA and AA. In GA it would be a completely new session if someone came from another email campaign. Thank you for the help, this is has been really useful.

[Jennifer_Dungan]

Yes, though you should also be aware that GA4 changed those behaviours....

In Universal Analytics, each new campaign would start a new session...
In GA4, new campaigns do not trigger new sessions...

Also, campaigns by default in GA4 are no longer set to a 6 month retention... (this could be changed, but that was the previous default). Now, in GA4 it says "Session Channel" so it makes me believe that campaigns are all now session based... but I haven't really tested it out... GA is our backup analytics, we rarely dig into the campaign traffic there.

Adobe sessions/visits are all based on time (default is 30 mins of inactivity), and then you have full control over your Marketing Channel and their retention, your Tracking Code retention, and any custom tracking that you choose to use.

[AlexanderRo4]

Thank you for that.

While implementing your solution for my segment I noticed the number of visits doesn't change whether it's 14 day or 30 day last touch attribution, it remains the same total.

This was the solution you provided.

HIT [

    campaign starts with arcataff

    AND

    VISIT [

        HIT [

            campaign starts with arcataff

            AND

            campaign instance exists

        ]

    ]

]

[Jennifer_Dungan]

The second solution I provided cuts out the attribution.. only the pages inside of a visit that set the campaign will be included....

If you care about the attribution... so visits within the 14 days (or 30) that come in direct after the initial campaign, you need to use the simpler segment.

[AlexanderRo4]

Actually the simpler segment still gives the same results for 14 day(or 30).

[Jennifer_Dungan]

Hmm.. I did the same in our data and there's definitely a difference between the two segments.

Looking at Last Month, on my test, I see:

• 31% difference on Page Views
• 25% difference on Visits
• and 1% difference on Unique Visitors

Note that my retention period is 7 days on Tracking Code.

Now are you applying a custom attribution model after this segment logic is run? All my details are about the attribution set on your campaign dimension... if you are superseding the logic with attribution that might be causing some oddities?

Or maybe your tracking is in a constant state of overriding values so you aren't ever getting the benefit of attribution... 

Without seeing your implementation this could be hard to diagnose....

[Jennifer_Dungan]

I should also ask, are you at least down to your proper campaign? If not, you may have missed a detail in the segment?

[AlexanderRo4]

The attribution model is set to the Visits metric itself. It's not set to the dimension.

[AlexanderRo4]

This is what I have set for tracking code.

[AlexanderRo4]

If you can give me some clarification on what you meant by this paragraph:

"Now are you applying a custom attribution model after this segment logic is run? All my details are about the attribution set on your campaign dimension... if you are superseding the logic with attribution that might be causing some oddities?"

[Jennifer_Dungan]

OK, good, so this is set on your actual dimension (1 Week Attribution)... I am not then sure why you are saying there's no difference between 14 Day and 30 Day Attribution.... are you applying custom attribution on top of the dimension attribution?

As for:

If you can give me some clarification on what you meant by this paragraph:

"Now are you applying a custom attribution model after this segment logic is run? All my details are about the attribution set on your campaign dimension... if you are superseding the logic with attribution that might be causing some oddities?"

---

I'm trying to understand what you are doing and seeing...  You've shown me that your dimension is set to a 1 Week Attribution. If you are then using the segment and applying it to a dimension with a custom attribution I have no idea what that is going to do... like cause issues with the calculations, since the segment is based on the dimension attribution (1 week), if you are then stacking that on a custom attribution model it very well might be overriding a lot of the logic of your segment.... I don't know, I haven't mixed attributions like that.... 

[AlexanderRo4]

This is with the segment.

And this is without the segment.

Am I doing this wrong in some way?

[Jennifer_Dungan]

OK, you are applying custom attribution on top...

I think in this case, the segment is forcing the attribution to default 1 week for all cases... so trying to apply 14 days and 30 day after the fact aren't going to work.

I think in this case, if the unsegmented table is showing something like:

• A
• B
• C
• D
• etc

And you just want to see "C", your best bet is probably to just apply a filter to your campaign dimension...

I see that you already have an active filter in your table...what is it doing and can it be modified to perform both duties?

The problem is that the segment is looking specifically for hits with arcataff and that does not exist beyond the week as per your dimension settings.

Without the segment, Adobe can do a calculation of the 14 day or 30 day attribution... but once you apply the segment it prevents those attributions from applying.

But if you just use the filters, you aren't locking the results to just hits with value, you're only limiting the list of items, but not changing the calculation.