Performance measurements… and the people who love them

This post clarifies how to measure web request latency by standardizing definitions (e.g., TTFB), mapping Catchpoint and browser timing phases, and exposing how common percentile aggregations can obscure tail behavior; by visualizing full distributions with CDF/QQ plots and modeling the impact of many concurrent page requests, it argues organizations should emphasize tail latency (p99+) and user-experienced latency over medians, demonstrating scenarios where Cloudflare is faster at high percentiles despite similar or higher average request latency, and previewing new tools to help customers see these effects.