If you have a lot of confidence and/or experience in what an ideal end state would be for your use case, then:

1. Be explicit about identifying that end state and what it would look like.
2. Incrementally, as-needed, solve the short term problems biasing for solutions that will get you closer to the ideal end state.

If you don’t have a lot of confidence and/or experience in what an ideal end state would be:

1. Ship incrementally till you gain confidence in what your ideal end state would be.