A-B testing, or split testing, lets research firms compare competing messages, concepts, or calls to action with real data instead of opinions.