Percentiles
Paste raw samples and get p50, p90, p95, p99 along with the mean, so you can see how far the tail is from the average.
Nearest-rank percentiles, the same definition most latency dashboards use.
Averages hide the thing you care about. A service averaging 90ms can be perfectly healthy or can be serving most requests in 40ms and a painful slice in 900ms, and only the percentiles tell you which. The gap between p50 and p99 is the number worth watching.
Paste numbers separated by anything at all — spaces, commas, newlines from a spreadsheet — and it sorts them, works out each percentile, and shows the spread. It is deliberately blunt: no sampling, no estimation, just the actual ordered values.
What people open it for
- Turning a column of latency measurements into the percentiles for a report
- Showing the distance between median and tail latency
- Checking a monitoring dashboard's numbers against the raw data
In the terminal
p99 12 40 91
Every utility is also a command in the desktop's shell.