Median & Percentiles

When measuring application performance, a common practice is to report the median together with percentiles (e.g., p50, p90, p99). The letter p denotes the percentile.

• Median (p50): Sort a set of request latencies from fastest (smallest) to slowest (largest). The value in the middle is the median, also called the 50th percentile (p50). This means 50% of requests are at or below this latency, and the other 50% are at or above it.
• p95, p99, etc.: The 95th percentile (p95) is the latency at or below which 95% of requests fall (only 5% are slower). Similarly, p99 covers 99% of requests.

Example:

If your p50 is 120 ms and your p95 is 480 ms, then half of requests are ≤ 120 ms, and 95% are ≤ 480 ms.