Preferred to the mean for salary data because one outsized package does not drag it around.
Still a single number describing a shape. Two cohorts with an identical median can have completely different spreads, and therefore completely different advice attached.