> ## Content Index
> Fetch the complete content index at: https://debugly.dev/llms.txt
> Use this file to discover other available public pages before exploring further.

# Your Average Latency Is Fine and Your Users Are Not
- URL: https://debugly.dev/p99-latency-and-why-averages-lie/
- Published: 2026-09-04T15:30:00.000Z
- Updated: 2026-10-23T09:12:42.000Z
- Description: Mean latency is a number that describes nobody. Here is why the tail dominates real user experience, how retries multiply it, and which percentiles are…
- Author: Rohit Bhadani
- Tags: Performance, Debugging, DevOps

| Percentile | What it tells you          | Who it describes                            |
| ---------- | -------------------------- | ------------------------------------------- |
| p50        | The typical request        | Your median user, one request               |
| p90        | Where the bulk ends        | Users on a slightly worse path              |
| p99        | The tail begins            | One in a hundred requests, every heavy user |
| p99.9      | Your worst real experience | Your most valuable and most active users    |

---

Related reading: [HTTP Keep-Alive vs TCP Keep-Alive: They Are Not the Same Header](https://debugly.dev/http-keep-alive-vs-tcp-keep-alive-comparison/) explores a connected angle.

Related reading: [Valkey-proxy: Migrating Redis Without the Rewrite](https://debugly.dev/valkey-proxy-redis-without-the-rewrite/) explores a connected angle.

Related reading: [The 4 MB JSON Response That Compression Fixed](https://debugly.dev/uncompressed-json-response/) explores a connected angle.

Published on [debugly.dev](https://debugly.dev/) by [Rohit Bhadani](https://debugly.dev/about/) | Founder of [Krova Cloud](https://krova.cloud/?ref=debugly.dev)