Lab data vs. CrUX field data: which gates your render budget

Lab data vs. CrUX field data: which gates your render budget

Lighthouse lab and CrUX field measure different things and disagree often. Compared LCP on 90 templates.

— Lab (Lighthouse, throttled): controlled, single-run, median 1.9s ▓▓░░░░░░░░
— Field (CrUX, p75 real users): median 3.4s ▓▓▓▓▓░░░░░

Delta median: +1.5s field-over-lab; p90 gap hit +4.2s on image-heavy templates.

So what: Google ranks on field p75, not your lab score. A green Lighthouse run on a fast machine is a debugging tool, not a verdict. Use lab data to iterate quickly (it's deterministic); use CrUX/Search Console CWV to know if you actually passed. If you only have one, trust the field number — it's the one with users and rankings attached.