Philip Walton on the difficulties the webperf community faces when discussing web performance. Simple numbers don’t cut it. We need to provide context when discussing performance results.
What concerns me about this practice is that it glosses over a lot of important nuance, and it perpetuates the idea that synthetic or lab-based tools (like Lighthouse, WebPageTest, and many others) are genuine and precise assessments of a site’s actual, real-world performance—rather than what they are: tools to test, debug, diagnose, optimize, and predict performance or detect regressions under a set of controlled conditions.
I’m definitely guilty of the simplicity he discusses. Thanks for the challenge Philip.