Well, that’s weird…
Spend long enough investigating web performance and you collect some pretty odd stories. I’ve investigated improvements that no one made and regressions that sent developers hunting through perfectly innocent deployments. Sometimes the issue was in the measurement itself; sometimes it was in another team’s idea of who should be visiting the site.
This talk is a collection of stories from real projects, with practical advice taken from the things that caught me out. We’ll look beyond headline percentiles, investigate bugs in performance tooling, and see how a change in visitors hid successful optimisation work. Along the way, I’ll share some simple ways to make comparisons more useful and experiments more trustworthy. Expect a few surprises, some pedantry, and lots of war stories.