You're designing an A/B test to evaluate the impact of showing content from non-friends in users' feeds. How would you test this with proper randomization?