Traffic Split in A/B

HardA/B Testing & Experimentation00:00

Practice interviewer

In session

5 left

00:00

Your question is Traffic Split in A/B. Take a moment with it on the right.

Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).

You need to log in / sign up to chat or submit.

Problem

How do you split traffic in an A/B test?

Asked in the Manager interview stage. Project follow-up.

Answer this as a practical design exercise. Explain how you choose the randomization unit, allocate users between control and treatment, ramp exposure safely, and verify that the split is working.

State the allocation you would start with and when you would change it.
Define a primary metric, guardrails, and an explicit MDE before launch.
Show how you would estimate required sample size and test duration.
Describe checks for sample ratio mismatch, peeking, and incorrect analysis units.
Give a clear ship, do-not-ship, or iterate rule.