A/B testing shows 50% of visitors version A and 50% version B, then measures which converts better. It's the gold standard for data-driven optimization — but only when you have enough traffic to reach statistical significance.
The traffic threshold most teams ignore: to detect a 10% relative improvement with 95% confidence, you typically need 1,000+ conversions per variant. If your page gets 500 visitors/month and converts at 3%, that's 15 conversions/month — you'd need to run the test for over a year. Below this threshold, A/B testing is a waste of time. Use analysis tools and best practices instead.
What to test first
Test the highest-leverage elements first: headline (biggest impact), CTA copy and placement (second biggest), hero section layout, and social proof positioning. Don't start with button colors — the "red vs. green button" debate is the least interesting question in CRO. Test the words, not the colors.
One variable at a time. If you change the headline, CTA, and hero image simultaneously, you won't know which change caused the result. Multivariate testing exists but requires even more traffic.