beginner · metrics

Why p99 and Not the Average?

Why do teams look at p99 latency instead of the average? And when is p99 the wrong number?

Practical scenario

A weekly report shows "average API latency: 120 ms, well within target". Support has been escalating complaints about slow page loads all week.

What it tests

  • Understanding that latency is a distribution, not a value
  • Whether the candidate can also argue against p99
  • Awareness of how traffic volume changes what a percentile means

Lessons behind this question