How is algorithmic randomisation improving insertion in binary search trees?

I know that randomisation in quick sort input may help into avoiding the worst case scenario which is an already sorted list. But how is this helping to improve into an insertion in a binary search tree?