I prefer CODE V for the main reason it gets designs done faster. The optimizer handles constraints more efficiently than others and therefore you spend less time with weights (although there will always be some weighting required), and targets are achieved quickly. You end up with a design form that is generally superior to other methods where the designer is required to manually massage the optimizer.
Tolerancing models are accurate (and extremely fast with wavefront differentials) meaning the as-built performance matches predictions so that redesign is not usually required.
The analyses algorithms generally have safeguards to prevent the user from getting wrong results. While the designer always needs to validate assumptions, this is less of an issue with CODE V than other packages, and reduces user error.
.
