"A good model enables you to be wrong faster.
In a review meeting where a decision keeps getting deferred, the room keeps asking for one more datapoint, as if it will remove responsibility. Meanwhile the calendar keeps committing you.
Uncertainty is not the enemy; implicitness is. You see it in a design decision that is legible today and costly tomorrow, where a missing boundary becomes a permission slip for everyone to improvise. If you introduce a measure, name the distortion it creates, because the distortion is part of the cost you are choosing.
Write the stop line now: what evidence ends the attempt, and what you will do when it appears. That is how you keep pressure from rewriting the decision. Keep it short enough to remember under stress.