This one is difficult to achieve but I think I produced the right values although the aesthetic result is not quite good. That is because I was not able to hide redundant marks without affecting data densification which would result in wrong results. So I switched off marks stacking but depending on the number of matching combinations they are layered on one another giving this ugly look as though they were not positioned precisely well.
The columns show what the first product is and the rows hold the consecutive product purchased. The [First Product] and [Not First Product] are LoD calculations and are then used as dimensions in cross table so I hid the row & column where these yield NULL (as we are not interested in such combinations).
This is what it look like for all the customers on their 4th order:
and here is a selection of customers:
As you can see the diagonal shows consecutive orders of the same product as the first one so you probably want to ignore those and I could not hide them for the aforementioned reason. I am not entirely sure if the values are right as I only checked a very few situations.
consecutive order 10.0.twbx 50.7 KB
I think it is very close. We have decided to send this off as part of a bigger piece of work with some true data nerds. Really appreciate the effort in your response.