You can have multiple discretes (blue pills) on the color shelf, so all you need to do is turn your aggregate measure into a Discrete and then add it to Color. However, what Tableau does is assign each of top Color discrete (by positioning on the Marks Card) discrete values to a separate color in the default color palette, and then the group of distinct values of the lower-level discretes on Color is used to create a range of colors. However, whenever one of those discrete measures changes value, you'd end up with a new color being assigned and would have to manually edit the color. For that reason, I suggest instead of just using the number of records, you use some sort of binning algorithm.
I did a presentation that goes into more detail on this for the Atlanta Tableau User Group here's a link: http://community.tableau.com/thread/126047.
Thanks, but I don't understand. I got as far as putting two discretes on the colour shelf (see attached), and reassigning the colours for Elapsed Days to red, orange and green. I can't see how to get gradations of colour based on No. of applications, although maybe it's covered in your explanation, which I don't entirely understand. I've looked through your Atlanta presentation and I'm still stuck.
Can you shed any further light? Perhaps I would be better off using size or some other attribute to call out the number of applications?
Anyway, I really appreciate your help here.
Just to be clear, are you looking for individual bars to have gradients within the bar or do you want the bars with a higher number of elapsed days to have a different color/gradient? In the former case (gradients within the bar) Tableau doesn't support that, though during the v8 beta I found a way if your data has enough level of detail and you're willing to draw 1000s of marks. But I wouldn't recommend that.
And for more clarification, here's another way to describe what Tableau is doing with multiple discretes on the Color Shelf: Tableau is *not* applying a gradient across _all_ colors of the Elapsed Days dimension based on the range of values of the No. of applications measure, Tableau is only generating a "gradient" (really a range of categorical colors) for the set of distinct values of the child discrete pills (No. of Applications measure) for each value of the top-level discrete pills (Elapsed Days dimension).
In the latter case (which is what I was describing), the thing you're running into is that the only dimension in either view is Elapsed Days, so there's only one value of the No. of Applications measure for each value of the dimension. What happens when Tableau does the color assignment is that the Elapsed Days values are used to assign the color range (and get one color each), and then when you add No. of Applications there's only one value for each Elapsed Days, so there's still only the one color.
If you filter your data such that the No. of Applications changes, you will see Tableau assign new colors because the combination of Elapsed Days & No. of Applications has a new value. You could manually assign the right color, but every new combination of values would require yet another manual color assignment. That's why I was suggesting a binning calculation like IF SUM([No. of applications]) > 250 THEN "expected" ELSE "high" END, then you could put that as the 2nd discrete pill on the Color Shelf. You'd have to either play with filter settings to make sure that you are able to generate the different combinations of Elapsed Days and the new measure so that Tableau would spit out the new colors, which you could then edit and fix in place. This is because Tableau doesn't pre-compute the domain of the measure (i.e. all of the potential values), so we have to play with the data to get Tableau to display them.
Does that help?
It's a lot to take in, but I think I understand and I'm going to experiment with your suggestions.