Core Web Vitals: measure before you optimise

Separate lab and field data

Lighthouse diagnoses controlled test conditions; field data reflects different devices and networks. Good thresholds are LCP at most 2.5 seconds, INP at most 200 milliseconds and CLS at most 0.1, measured at the 75th percentile.

Find the actual bottleneck

For LCP, inspect response time, the main image, fonts and render-blocking resources. Reserve image and content dimensions to reduce layout shifts.

Reduce interaction work

Inspect long tasks and main-thread JavaScript. Break up expensive work, avoid unnecessary hydration and test actual clicks and form input, not only initial loading.

Verify after changes

Compare before and after under consistent conditions, then monitor field data when sufficient. Good scores do not guarantee rankings or sales, but they reduce usability friction.