Another brilliant talk from one of the older goto conferences: Why Scaling Agile Doesn’t Work.
My main takeaway from it: instead of writing user-stories try writing hypotheses:
Why is this better – because of the very last part. You clearly write how are you going to measure the effect of the feature – and it’s stated in the very first document you write. So this feels like you are writing a unit-test but not for the code but rather for your user-story.
And the signal may be pretty simple in most cases – just test how many times your feature was used. If the answer is “zero” – your feature is useless, turn it off (oh, you have to build feature flags for that of course).