Most async design reviews fail for the same reason: the post asks for feedback in general instead of feedback on something in particular. Everyone answers a slightly different question, the thread grows to forty comments, and the designer is left to guess which opinion carried weight.
Say what has already been decided
Open with the constraints. Which flows are locked, which platform is first, what shipped last quarter that this has to sit next to. Reviewers who know the boundaries stop relitigating them, and you stop reading the same objection three times.
Ask exactly two questions
Two is the number that works. One is usually too narrow to catch a real problem, and four invites people to answer only the easiest one. Write them as questions a reviewer can answer with a position, not with a shrug: does the empty state carry enough of the value, or does the pricing step belong before the workspace step.
Name who decides
Every review needs one person who will close it. Not a committee, not a vote. Put the name in the post so contributors know they are advising rather than approving, and so nobody waits three days for a decision that was never assigned.
We started doing this in Cloudhub eighteen months ago. Review threads dropped from an average of thirty-one comments to nine, and the median time from posting to a decision went from four days to one.
