The approach I find most helpful is to consider whether the feature will help make progress toward our target outcome. If so keep considering it. If not, dump it.
Then from there, consider whether the feature helps to address risks or validate assumptions. If so, you may want to work on it sooner rather than later. It’s also helpful to consider if technical implementation considerations drive doing one feature sooner than others.
This approach works best when you’re considering several features that are all associated with the same general effort.