Core Web Vitals.

Core Web Vitals is the collective name for the three metrics Google defined to measure a page’s real user experience.

The three metrics measure three separate dimensions and none substitutes for another: loading speed, interaction responsiveness and visual stability. For a page to count as good, all three have to pass the threshold.

1. LCP — the paint time of the largest visible content. Good threshold: 2.5 seconds.
2. INP — the visual response time after an interaction. Good threshold: 200 milliseconds.
3. CLS — unexpected layout shift. Good threshold: 0.1.

All three are assessed in field data at the 75th percentile — the value experienced by three quarters of users. That keeps a single bad sample from spoiling the score while letting a widespread problem show.

The set of metrics changes over time: INP replaced FID in 2024, and new metrics may be added in future. So improving the experience a metric measures is a more durable approach than optimising for the metric itself.

The data used in ranking is field data, not a lab score. Getting a high score in an auditing tool does not mean being good in Core Web Vitals terms.

The three metrics are independent of one another: a page can open very quickly and respond late to clicks, or be fast and responsive and still shift under you while loading. Each has its own treatment, and the work that fixes LCP usually does not touch INP.

Its weight in ranking is limited, and Google says so plainly: content relevance always comes first. Poor Core Web Vitals do not bury a good page, but they can make the difference between two equivalent ones.

The real justification is not ranking but conversion: a slow, unstable page loses the user regardless of ranking.

1. If LCP is poor — look at resource size, discovery delay and render blockers.
2. If INP is poor — take an inventory of JavaScript, break up long tasks.
3. If CLS is poor — dimension reservation, font matching and content arriving later.
4. If all three are poor — look at the architectural decision; optimising one by one will not do.
5. If all three are good — the purpose of measurement is no longer improvement but guarding against regression.

The last item is a way of working: once the metrics fall below the threshold the job is not over, protection begins — and protection only works if it is tied to a test.