Expand my Community achievements bar.

Join us for the next Community Q&A Coffee Break on Tuesday April 23, 2024 with Eric Matisoff, Principal Evangelist, Analytics & Data Science, who will join us to discuss all the big news and announcements from Summit 2024!
SOLVED

creating segments for revisitors during two different date ranges

Avatar

Level 2

Hi!

We run a campaign a couple of times a year, and I'm curious about the repeat customers who keep coming back and engaging (visiting and ordering). The problem is, when I try to build segments for these loyal visitors, it keeps grabbing folks who only participated in the second campaign. I've tried excluding them, but no dice.

 

Creating the desired segment:

Let's say I want a segment of people who visited a specific page between January and June, ordered something then, and returned for another purchase between July and December. I used the "after sequence" feature to track July-December orders, but as you guessed, it included folks who first bought in the second half. Ideally, I only want customers with a purchase history in the first half.

I tried excluding those cases by adding visit number or customer loyalty dimensions to the bottom visit container and excluding them. Unfortunately, that ended up wiping out most of the results (probably because the exclusion was in the visitor bucket. visit or hit buckets did not changed the results).

스크린샷 2024-01-12 102657.png

 

how can we do this?

1 Accepted Solution

Avatar

Correct answer by
Level 2

In my recent meeting with Adobe, we discussed inconsistencies in traffic data. Orders and visits from seemingly new visitors were categorized as low traffic due to discrepancies between backend and UI-displayed visitor IDs. Essentially, different backend IDs might be getting grouped under the same ID in Workspace. It is better to use a data warehouse when we look into visitor IDs

View solution in original post

15 Replies

Avatar

Community Advisor

Did you try involving Experience Cloud ID or MID in the calculation?

Avatar

Level 2

For the initial assessment of my segment's performance, I opted for the visitor ID dimension as I'm still familiarizing myself with Experience Cloud ID and MID

Avatar

Community Advisor

Hmm while that should work, I was having some issues replicating as well... but instead of using a date range, I added a sub container inside the first clause for: 

 

Month equals Jan 2023

OR

Month equals Feb 2023

OR

Month equals Mar 2023

OR

Month equals Apr 2023

OR

Month equals Jun 2023

 

 

And then in the second part, another container for the Months Jul - Dec

 

 

Jennifer_Dungan_0-1705034251125.png

 

 

Try this and see if it works better?

Avatar

Level 2

To clarify, the 'test seg' I mentioned earlier is the one with the specific structure we discussed. The 'V2 seg' is the one that uses a sub-container with the 'or' condition.

스크린샷 2024-01-12 172720.png

The example I'm showing you now falls under the category of visitors who only made purchases in the second half of the period

스크린샷 2024-01-12 172442.png

Avatar

Community Advisor

Keep in mind that the "ONLY After Sequence will not show the purchases in the first half.. only the second.

 

When I tested, I had ALL purchases (unfiltered), next to the segment filtered content.

 

I could see before and after purchases in the all orders column and only the second half purchases in the segment column...

 

In the date range one, I had people who didn't trigger the event in the first half... but that should have worked.. I think they may be making tweaks to the date logic again... 

 

Avatar

Community Advisor

If you don't include the "only after sequence" then you will see the purchases in both tests...

Avatar

Level 2

I understand that point, which is precisely why I shared the image. It demonstrates the metric values without any segmentation applied. As you can see, the visitor ID in the image didn't make any purchases during the first half of the period, yet they're still included in the "only after sequence" segment results.

Avatar

Community Advisor

I just recreated my samples again (using a different metric than orders), and both the Segments created with "date ranges" and with "specific months" return the same data, and it looks correct....

 

The issue is the overall "Visits" column is returning all visitors in conjunction with the segments, making it look like they are pulling back more Visitors than they should... 

 

You grabbed a Visitor that was returned due to the general "Visits" and not based on the segments.

 

Jennifer_Dungan_0-1705289976814.png

 

 

And if I sort low to high, which should be "0", if there were visitors returned in the first part that shouldn't be, you can see that there is at least 1:

 

Jennifer_Dungan_1-1705290099804.png

 

 

Here is what happens if I add a general Visits column:

Jennifer_Dungan_2-1705290244533.png

 

 

Your segment is probably working, it's just other columns in your table are making it look incorrect.

Avatar

Community Advisor

For reference, here were my final segments, I am giving you the "only after" ones, as they are the exact same but "Everyone":

 

Date Range Version:

Jennifer_Dungan_1-1705290662207.png

 

 

 

Specific Month Version:

Jennifer_Dungan_0-1705290643147.png

 

 

Avatar

Community Advisor

Actually, I am going to apologize.. I realized I forgot to not just check the visits, but to check the actual metric as well.. so I went back to take additional screenshots, and that isn't working... it should but it's not... 

 

I think Adobe has been tinkering with the logic behind segments, because stuff like this used to work... and by the logic of the the AND it should only return hits that have both criteria, and it's not...

 

I will be meeting with someone on the Adobe team who works with the Segment builder, probably in the next month or two.. I don't have a date yet, since I am working on a project for them, I will bring this up... 

 

But at the moment, you're right, this isn't working as expected, and I don't know how to fix it, since the segments are built the way they should be built to accomplish this.

Avatar

Level 2

Thanks a bunch for your continued support and for dedicating your time, even though this isn't directly related to your work. I really appreciate it!

Just to double-check that we're on the same page, I compared the visit and order counts for segments with 'only after', 'only before', and 'include everyone'. Interestingly, some visitor IDs showed the same number of occurrences in both the 'only after' and 'include everyone' segments, while the 'only before' value was always 0.

When I dug deeper into these specific cases, I found that the visitors didn't have any orders from the second half of the period. This confirmed that their orders weren't omitted by the 'only after' segment; they simply didn't make any purchases during that timeframe and only ordered in the second half.

 

I'll keep digging and let you know if I find the answer.

Avatar

Community Advisor

Well if the first clause is failing, then it stands to reason that the second clause could be failing... it's like its returning the visitors who made an order in either or both time frames. Even though I specifically set the container to HIT... that hit should be both within the time or months specified and the specific action should have been performed. The THEN clause should ensure that both parts need to pass to return the visitor...

 

And while this may not be directly related to my work, you never know when I may have to build a similar segment... helping you is like building a repository of ideas for later.

 

Update: It looks like its ignoring the "Event" completely? I have visitors that don't have the event at all being returned... I took one of the Visitors that were part of the data set returned, and they have never triggered that event (so far as I can look back that is).. unless Adobe has some backend stitching between ECIDs so that our "Visitor ID" doesn't line up with the UV definition... thereby making it appear that the segments are failing? It's a small enough subset of data that this might explain the discrepancy... the user I am checking has only hit the site with 3 occurrences ever, and they were for a subscription flow.. making me suspect this was a QA user testing a flow in prod... 

Avatar

Level 2

In my scenario, I needed to analyze a specific page related to orders, even though it's typically an event page where ordering isn't possible. This required further segmentation of visits using sub-containers. Interestingly, unlike your situation where visitors without any event are included in the segment, I haven't encountered such cases yet.

Avatar

Community Advisor

Right, I just simplified my examples... I could have also added a specific page to it.. but if this doesn't work... then adding a page is likely to fail as well.

Avatar

Correct answer by
Level 2

In my recent meeting with Adobe, we discussed inconsistencies in traffic data. Orders and visits from seemingly new visitors were categorized as low traffic due to discrepancies between backend and UI-displayed visitor IDs. Essentially, different backend IDs might be getting grouped under the same ID in Workspace. It is better to use a data warehouse when we look into visitor IDs