See the attached. There are a couple of complexities to this solution. First of all, Tableau does not partition table calculations on the Measure Names convenience dimension, so we have to un-pivot the data so there is a dimension to identify the measure. I used Custom SQL to do this, you could use something else. (In general, Tableau works better with "tall" data rather than "wide data).
The second part is that Tableau tries to make life easier by automatically padding out data when we use a table calculation with a compute using of the date. Most of the time this works in our favor, this is one of those times when it doesn't. So, rather than using the date I used a string field based on the date.
Hope this helps!
Last filter.twbx.zip 71.9 KB