When testing news ads, test the forests first, then the trees, then the branches, then the leaves. In other words, test radically different approaches (the trees) unless the same approach has been your control for years - in which case it makes sense to test smaller changes (the leaves).