PageSpeed Insights vs real-user data: chasing the wrong 100

PageSpeed Insights vs real-user data: chasing the wrong 100

Month 1: My DIY site scored 64 on PageSpeed Insights mobile. I became obsessed. Lazy-loaded everything, deferred every script, inlined critical CSS. The number climbed.

Month 3: I hit a glorious 98. I celebrated. Then I opened Search Console's Core Web Vitals report — the field data, real Chrome users — and 40% of my mobile URLs were still flagged "poor."

The disconnect: PageSpeed's lab score runs on a simulated fast connection. My actual audience was on mid-range Android phones on patchy connections, where my heavy hero image still choked LCP. I'd optimized for the lab, not the people.

Month 5: I trusted the field data instead. The fix wasn't more script-deferring — it was serving smaller images to slow connections and killing one third-party widget. Real LCP dropped from 4.1s to 2.2s.

Month 7: rankings on mobile-heavy queries rose noticeably. The lab score barely moved — it was already fine.

The takeaway: PageSpeed's lab number is a diagnostic, not the goal. Google ranks on field data — real users on real phones. Chase the Search Console Core Web Vitals report, not the green 100. They measure different worlds.