Why your Lighthouse score disagrees with everyone else's
Two tools measuring the same page routinely disagree by thirty points. Almost always it is throttling, cold cache, or lab versus field data — not a bug.
By Sapun Lamichhane · Arcetis
You run PageSpeed Insights and get 42. You run Lighthouse in DevTools and get 78. You run a third tool and get 91. Nothing is broken; you are looking at three different measurements presented with identical-looking numbers.
Lab data is a simulation, not a reading
A lab score is one page load, on one machine, under a simulated network and CPU throttle. Change any of those and the number moves legitimately. The three variables that account for most of the disagreement:
- 1CPU throttling. PageSpeed Insights simulates a mid-tier mobile device. Your laptop is not one, and an unthrottled desktop run will always look dramatically better.
- 2Network shaping. Simulated 4G with added latency versus your actual connection changes LCP more than most code fixes will.
- 3Cache state. A warm cache produces a fast run that tells you nothing about a first-time visitor, who is the person the score is about.
Field data measures something else entirely
The Core Web Vitals in Search Console come from the Chrome User Experience Report: real visits, real devices, real networks, aggregated over 28 days. That is what affects your ranking. A lab score is a diagnostic tool for finding causes — it is not the number Google uses.
Lab data tells you why a page is slow. Field data tells you whether it is slow for the people actually visiting.
What to do about it
- Compare like with like: always the same tool, same throttling profile, same cache state, before and after a change.
- Treat a single run as noise. Three runs and a median is the minimum honest comparison.
- Fix causes, not scores. Render-blocking scripts, unsized images and oversized JavaScript bundles move both lab and field numbers; chasing the score itself moves neither reliably.
Our Performance checker runs a real Lighthouse pass against the live page and reports the measured values alongside Google's published thresholds — plus the page-weight and render-blocking issues that usually explain the number, which is the part a bare score never gives you.