Lab scores are a flashlight. Field data is the report card.
Free for 3 URLs.
People screenshot a 98 on PageSpeed Insights and ship. A week later Search Console still says LCP is poor. Both can be true. The 98 was a lab test with a warm-ish cache and no real users. The Search Console number is 28 days of Chrome users.
Lab
Controlled. Repeatable. Good for "did this image preload help?". Bad for "are we passing in Manchester on 4G?".
Field / CrUX
Real page views, aggregated, delayed, grouped by origin or URL when there is enough traffic. This is what page experience uses. Low-traffic URLs may have no URL-level field data. Then you watch the origin and your own RUM if you have it.
A sane workflow
- Use field data to decide whether you have a problem.
- Use lab traces to find the LCP element and the long task.
- Ship a fix.
- Watch field p75 over the next 2 to 4 weeks, because CrUX is slow to move.
Daily monitoring sits in the middle: it tells you the day a deploy made lab LCP worse, before you wait a month for CrUX.
FAQ
Why is my PSI field section empty?
Not enough Chrome users in the CrUX sample for that URL. Measure the origin, or collect your own vitals.
Should I chase a 100 lab score?
No. Chase LCP under 2.5s, INP under 200ms, CLS under 0.1 in the field.