A couple of questions for you. Do you have a compelling reason to be using the ACS microdata (PUMs files) like the need for custom crosstabs? Or, are you needing exact medians based on the microdata sample as opposed to the published estimates of medians in the ACS summary data? I have posted several census tract level ACS datasets pre-loaded into Tableau based on the 2010-2014 ACS summary data estimates, that is, five years of ACS sample data summarized to standard geographic layers. I prefer census tracts because the ACS provides reasonable estimates at that level and Tableau can automatically roll that up to county and state, for example.
My free ACS data is posted at here at my website www.thedemographics.com under “Demographics for Tableau.”
The census tract data uses a Mapbox background map but you can only create symbol maps based on census tract centroid points not a filled map. As to your first question, as yet Tableau can’t join polygons to form a new, higher level geographic layer, although I think that functionality is in the works. So, you can only show Zip Codes color-coded to PUMAs as you have done, not actually spatially joined.
Hope that helps,
PS – I’m interested in any comments people may have about the usefulness (or not) of pre-loaded ACS data at census tract level. Thanks!