Thanks for the question! 
I would say that before releasing a new feature or an improvement, I know that we have something when I’ve heard about the same problem over and over again, we’ve put an initial solution (staging/prototype, etc.) in front of customers, and they react as though we have reinvented sliced bread. The excitement is obvious.
This does not mean that you will have a successful feature. It just means that you are on the right path.
After release, the only strong indicator I use is engagement. Are people using the thing?
We have released many features that had massive excitement at the testing stage and when released were never used or were barely used.
If the initial excitement is there but the post-release engagement is not, perhaps you just need to step back and figure out a more elegant, practical, simple solution, then do it all over again:
- Talk to customers.
- Show them initial solutions.
- Check excitement.
- Release the first version.
- Check engagement, and if not positive,
- Go back to step one.
At EnjoyHQ, we had a couple of features that became successful after 5+ iterations.
Don’t kill stuff just because it didn’t work the first time around. If the problem persists, you still have work to do. That’s how I go about it.