please explain use of set_output_delay

I personally understand it like this:

if we are analyzing the setup constraint in reference to a launch edge, then data has to be there before the setup delay constraint, so the setup constraint value will be added to the data propagation delay and that's the output delay maximum.

now if we are looking at the hold constraint then data will have to remain there after the launch edge, so the hold value will be cut out from the data propagation delay.

Everyone, please correct me if i am mistaking!!