4. Performance Tools
What to run, and what each one actually tells you.
These overlap, but each is best at one job.
Start with PageSpeed Insights for an unfamiliar page — it gives you lab and field data side by side.
Use WebPageTest when you know the page is slow and need to know why.
Do the work in DevTools.
Lab and field data
Worth understanding before reading any tool's output.
Lab data comes from a synthetic test: one machine, one network profile, one run. Reproducible, so it is good for catching regressions.
Field data comes from real users on real devices. Noisy and slow to update, but it is the truth.
Trust field data when the two disagree. A page that scores 98 in Lighthouse and p75 of four seconds in the field is a slow page.
Test on a real device
The most misleading thing you can do is test on your development machine. A recent laptop on office wifi is nothing like a three-year-old Android phone on a congested network, and the gap is widest exactly where it hurts — JavaScript parse and execute time.
Throttle deliberately: 4–6× CPU slowdown approximates a mid-range phone, and "Slow 4G" is a reasonable network default.