Hi @VincentF.
you are poking into the vasp nest here - this is currently not possible with vanilla pivot table functionality. I am (selfishly 
) pinning my thread on enhancements to unleash the potential of pivot tables here where I propose one feature exactly solving your case at hand and where I would love to get PD input on.
Currently, we are solving this by pre-aggregate data to specific defined dimensions (like region, account, product) per month and join the “last year” (current year month -12 months) sales as individual column. This way you can set up your pivot table formula for YoY Growth = sum(sales) - sum(sales-1yr) which will look like this:
Since your data is at daily aggregate, you would self-join in two columns, one for sales_-30 days (Month over Month growth) and sales -365 days (Year over Year Growth).
Note: with derived properties, I believe it is possible to aggregate the last year sales on the fly in workshop. However, you still have to to pre-aggregate your data, so that you have the fixed monthly/quarterly/yearly sum and the pk_last_year for the search around to aggregate the sales_-1yr. You could not just have your backing data to be each individual sales transaction…
Hope this helps!