How is traffic typically divided in an A/B test?

• A. 60% to variant A and 40% to variant B
• B. 50% to each variant
• C. 70% to the preferred variant
• D. 100% to the control variant

Answer: (B)

Traffic is typically split evenly between the two variants to ensure a fair and efficient comparison. When each version gets half the users, the sample sizes are the same, making the difference in performance more reliable and the test statistically powerful for the total number of participants you’re testing with. Skewing the split to something like 60/40 or 70/30 reduces precision because one variant accumulates more data than the other, which can bias the estimate and complicate interpretation. Giving all traffic to one variant eliminates the comparison entirely, so the test becomes unusable. Uneven splits are only used in special cases with a specific justification, but the standard, most effective approach is a 50/50 division.