Comment by jacquesm

One thing I've learned from A/B testing large sites is that you change only one metric at the time, but the domain clouding one sounds like an excellent candidate as well.