Measure delivery-time quality — mean, or the tail?

Metric designMedium

Problem. You're choosing the metric for a delivery-time experiment. Why is mean delivery time (or mean ETA error) a poor choice, and how do you handle the metric statistically?

Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.

← Design a switchback test for a real-time dispatch changeA change that nets to zero but helps some, hurts others →