Building a sequential segment

[KaterynaOs]

Hi,

I want to consult about building a sequential segment. I need to investigate "customer migration" from one brand to other brands.

We sell brand A on our site since Nov2023. I need to segment for customers, who bought brand A once, and then they did not buy brand A anymore during 6 months, however they bought some other brands. I need to know which other brands did they buy during these 6 months.

I created a segment: 

However, when I tested it on couple of individual customer IDs, it did not work. I see orders with brand A during the 6 months for the same customer.

What am I missing here?

Thank you in advance,

Kate

[igupta]

Hi @KaterynaOs ,

Very first thing I want to bring into your attention is that the segment needs to be visitor level as you are looking at the context of Visitors and this will not hold true for visits. I tried to build your segment with and without exclusion and I am getting 100% overlap in venn. Maybe try with visitor and see if this works?

[MandyGeorge]

So there’s a few things with your segment here. First, as was already mentioned, it does need to be at a visitor level, not visit. The level that you put your segment at determines where it’s going to look for the conditions and what it’s going to return. A visit container means that everything that happens has to be in the same visit. Since you want at least six months, it has to be visitor, no visit is going to last six months.

The second thing is more of just something to keep in mind. If a customer refreshes their cookies or uses a different browser/device, they are going to appear to be a different customer. So while a lot of visitors will still be using the same ID, it is possible that some aren’t going to be captured because they’ve reset their cookies or changed devices.

Next, the actual sequential segment. If I understand you want the conditions:
bought brand A once
bought not-brand A in six months

The actual logic gets a bit complicated. You want a visit with a hit where they bought brand A. Then you want a purchase without brand A, so you need two conditions, one of buying non-brand A and one of excluding anyone who does buy A. I think that the logic that you want will be something like this. 

The exclude logic should be at a visitor level, not hit or visit. If you're using a hit exclusion, you're just excluding the individual hit that had brand A, meaning the visitor could have still bought brand A and it would bring in everything else. But if you exclude at a visitor level, it means that there was no brand A purchase at all.

@Jennifer_Dungan is the real segment expert, perhaps she can weigh in too about the logic below.

[Jennifer_Dungan]

@MandyGeorge is correct, since you are looking at the Visitor data, the segment should also be Visitor Scoped. Note that even when looking at Visitors, the data is constrained by your Panel's data range... so in this case, since you want to look at 6 months, you will need to have at least 6 months in your range... now, I say "at least" for a reason....

Your sequence is looking for "within 6 months" to perform another action, but in theory, if you were looking at all of "Last Year" for example, users who did "A" in Jan, then "B" within the first half of the year would be included, and someone who did "A" in March could still be included if they did "B" in July, or August, etc.

I also agree with the exclude on the Visitor level.

I don't have any similar situation that I can test this particular logic.. but I would recommend trying to test something in a controlled environment (Dev or QA) over a few days (substituting 6 months for a day or two), and trying various scenarios.

Try every combination you can think of, and make sure you know which users are supposed to appear, and which ones aren't.... make sure your segment is returning the correct data.

[KaterynaOs]

Hi! Thank you both @Jennifer_Dungan and @MandyGeorge for your ideas. 

I will try to test it as you suggested with shorter periods.

Best regards,

Kate