You'll design a better system if you continuously evaluate it from at least two different viewpoints.

Something that looks "just right" from one perspective may be totally wrong from another.

I wonder how much of the benefit of formal methods, TDD, and testing in general can be explained by this.