See the attached for an example using Superstore Sales, where a calculated field called "Sales Bins" uses an IF/THEN statement to create the bins, and Order ID is the stand-in for Member ID. The blue pill in the view has "Ignore in Table Calculations" turned off. There's then a table calc called "Total Sales" that returns a single value for the total sales, using WINDOW_SUM instead of TOTAL because TOTAL won't function when we're using an aggregated discrete value (the Sales Bin) for partitioning. Then the Total Sales uses Order ID for the Compute Using. We move the Order ID onto the Level of Detail Shelf to remove it from the display but leave it in the view. Finally, there's a duplicated instance of Total Sales on the Filters Shelf to only show non-Null values, taking the view from 5649 marks down to 3. This can improve performance for larger data sets.
One correction to terminology: Technically, we are addressing (setting the Compute Using) on the Order ID/Member ID, and partitioning on the Sales Bin. The calculation is being performed for each value of Member ID in the bin (the addressing), and restarting for each new values of Sales Bin (the partitioning).
The sample you sent is perfect. It took a little work to figure out all of the "levers" to pull, but I've got a nice visual that works well.