The color condition you have set up is not a dimension, instead it is a discrete (blue pill) measure. See http://www.theinformationlab.co.uk/2011/09/23/blue-things-and-green-things/ for a more detailed explanation.
This gets tricky because the KPI color is based on a table calculation, so the view needs to have enough detail to enable that computation (and then any following computations) to occur. And I'm presuming not every KPI color for every measure is guaranteed to have the range of Red/Yellow/Green values every time, which means that we have to do our own "padding" to make sure that the values come about.
I set this up for one measure (Sales) in the attached, where there are three nested table calculations that get the store counts for Red/Yellow/Green, and then Measure Names can be used to generate the highlight table. I also added two table calculation filters, one to reduce marks and the other to hide the 2012 data. Using this technique, you'd need to generate an additional worksheet for the % measures, and then if desired 2 additional worksheets for other measures. Then you could do things like hide headers and put the worksheets together within a horizontal layout container in a dashboard. I don't know your data volumes, this could get pretty slow in terms of update times.
Alternatively, I'd be tempted to create this view by first working with the datasource to do more pre-computation, for example getting to the point where the YoY computation is done using a a regular aggregate or even done in the data source. Then there could be some scaffolding done to create false "Measure Names" that the results could be plugged into. Joe Mako did a Think Data Thursday on building scaffold sources a few weeks back: http://community.tableau.com/thread/126855.
That said, this could be a lot easier to create with something like this idea: http://community.tableau.com/ideas/1604.
counting KPI results.twbx.zip 111.4 KB