API Testing · Senior · Question 9 of 75
Why report p95 and p99 latency rather than averages in API load tests?
- 1Percentiles always look better, so reports are nicer
- 2Averages hide the slow tail that many users still hit
- 3Averages are impossible to compute for HTTP traffic
Show the answer
Correct answer: Averages hide the slow tail that many users still hit
Explanation
A 100 ms average can hide 2-second p99s. In microservices, one request fans out to many calls, so tail latency adds up.