Tag: experiment-design
Interview Questions
- Why you run an A/B test for whole weeks, not "until it's ready"
- When treating some users secretly changes the others
- Proving the cheaper system is "no worse" than the old one
- The feature that wins in week one and fades by week four
- One metric to decide, several to keep you honest
- Blocking on a strong predictor to sharpen your test
- Paired versus two-sample t-test, when to pair