You make it sound like you’re just shifting the problem elsewhere, when this isn’t the case.
It’s not wholly subjective. Some of the processes you can use to understand your data are mathematically proven. Many others are well-tested.
In any case the idea is to try to minimise your biases and check whether your assumptions are valid so that you can make better, more reliable, more informed decisions. It doesn’t have to be a perfect system to be better.
a) is this data accurate
b) is this data complete
c) is this data relevant
etc.
So even the act of selecting data is subject to bias, good judgment.