How would he know what distribution to use for the simulation without a sufficiently large dataset to use as an example? A simulation of a made-up system is pretty worthless.