One option (I think) is that you can unpivot your columns so that you end up 2 columns that contain the original column name (attribute) and the corresponding value. Then, you could put the column name column as an axis and the values as values and you would get the aggregation (i.e. count, sum, avg etc.) of the 3 textual values for each point on the axis. This could be a stacked or regular column chart, or 100% stacked if you want.
The thing is that I can't unpivot them, as they don't show up in my "Edit query" screen. I've written them all, using "SWITCH", from columns with numeric values. The question is basically if I can get more than one column in a chart. I have no problems making one like this.
What I want to acheive is getting more than one bar in the chart, as I have 11 that I need to visualize in the same space.
In your current dataset, since you have one column for each category, you can't have it show all categories in a 100% stacked bar chart.
For your requirement, you have to change your dataset into 2 rows, one for category (Purchase, Check-out, etc), the other column for result (satisfied, unsatisfied, etc). Then you put the Category column on X-axis, put Result coumn on Legend. Create a measure like "COUNTROWS()" as Values. It will return your expected result.