Find my approach as reference below and stored in attached workbook version 9.3
Configured as depicted below...Customer State on "Detail" and Customer Segment (your Business) on "Color"
trellis example.twbx 1.1 MB
This is helpful. However, when I try to recreate it; I seem to be getting multiple locations in each cell. Attached is a sample workbook. It may be that the shape of the data is different in your example vs mine.
Any help here is much appreciated.
trellis issue.twbx 58.5 KB
Checked your Workbook in version 10.1. I provided my workbook in version 9.3 and it worked perfect. I replicated the scenario once again in Tableau 10.0 and I must confirm your findings. Why version 10.0 is "acting" like this to my opinion strange. I will check again later today and come back to you.
Have a nice one.
Good morning Mahfooj,
Thanks for stepping in. Installed version 10 on my laptop and checked your provided workbook but the issue remains also in your workbook.
The issue is that multiple Locations are mentioned in one trelli where Sanjiv expects only the info for one location separated by applicable businesses. In version 9.3 it works perfect but in version 10.0 it goes "wild" and I can't get my head around WHY?;)
The main problem we've got with trellis
is that the datasource is not domain complete --
not all combinations of dimension values are present.
In the monthly view of the trellis example by Norbert
there is no sales in Alabama in October 2007,
so Arizona "takes a poll position" in the 1st trellis.
Other states are shifted accordingly, and so on.
More "holes" -- more "shifts" along the way.
Simply changing a Line for an Area would solve the problem.
The data densification behaviour is different for each chart type.
For short, Area charts behave more like Stacked Bars,
but Lines are different beasts :-)
The issue exaggerates in the trellis issue workbook by sanjiv.
Here we have the whole panes moving, since not every Business
happened in every Location (Months aside).
The common solution to the problem described above is
to have a complete domain of all Dimensions on a view.
In this particular case, the simplest approach looks obvious --
a plain old Data Blending using a "Scaffold" datasource.
Please find the attached.
Hope it could help.
Thanks for jumping in and getting the "Beast" on the table. Much appreciated:)
1 of 1 people found this helpful
Norbert, you're welcome.
Thank you for the response. This definitely helps clarify the issue for me. The idea of domain completeness, in this instance would be to have a scaffold, which is a cross product of all domains that can be bound against the original data source. It works perfectly.
Sanjiv, you're welcome.