Median and percentiles
The middle value of a list, and the values that mark off the slowest or biggest slice.
Line up a set of numbers from smallest to largest. The median is the one in the middle: half the values are below it and half above. It answers "what is typical?" better than the average when a few values are extreme. If nine people earn modest salaries and a tenth is a millionaire, the average salary jumps, but the median barely moves.
A percentile generalizes the same idea. The 90th percentile is the value that 90% of the data falls below. The median is simply the 50th percentile, often written p50. Pediatricians use percentiles on growth charts: a baby at the 75th percentile for height is taller than 75% of babies the same age.
In technology, percentiles are how teams talk about speed. Saying a website responds in 200 milliseconds on average hides the people who waited five seconds. So teams watch the p95 or p99: the response time that 95% or 99% of requests beat. If p99 is two seconds, one request in a hundred takes longer than that, and on a busy site that is a lot of annoyed people. Percentiles keep the unlucky tail visible.
Keep in mind
- The median is the middle value once the data is sorted.
- Unlike the average, the median ignores a few extreme values.
- The Nth percentile is the value N% of the data falls below; the median is p50.
- p95 and p99 are the standard way to measure slow requests.
