Percentile Latency
Also known as: p95, p99
Percentile latency (commonly p95 or p99) reports the response time below which a given percentage of requests fall — p95 of 200ms means 95% of requests were faster than 200ms. It's a far more useful performance metric than a simple average, which hides how bad the worst experiences actually are.
An average can look perfectly healthy while a meaningful slice of real users have a terrible experience — a system with a 50ms average response time can still have a p99 of 5 seconds if a small percentage of requests are badly slow, and the average simply buries that reality in the aggregate number.
This is why performance dashboards and SLAs increasingly specify targets in terms of p95 or p99, not average — "99% of requests complete within 300ms" is a meaningfully stronger, more honest commitment than "average response time is 100ms," which says almost nothing about the tail experience.
