Let's say the company MyTutors wants to test the effect that branding in the title tag has on search traffic.
It wants to test whether its landing page titles should be "A+ San Francisco Tutors | My Tutors" (Variation A) or "38 Tutors in San Francisco | MyTutors" (Variation B).
Split MyTutors landing pages into two groups.
Apply Variation A to the first group of landing pages and Variation B to the second group.
Compare search performance between both groups and declare the group with better performance the winner.
If the group with Variation B outperforms the group with Variation A, then we can make a data-driven SEO recommendation for Variation B's title format over Variation A's.