Core Web Vitals: the thresholds, where the data comes from, and what they are worth

Core Web Vitals is one of the few areas where the target is published as a number rather than described as a best practice. That makes it checkable, which makes it useful — and it is also routinely oversold as a ranking lever.

The three metrics and their thresholds

The gap between good and poor is the "needs improvement" band. The 75th percentile matters more than people expect: you are being assessed on the experience of your slower quarter of visits, not on your median. A fast page on a fast device does not exempt you from a slow one on a mid-range phone on mobile data.

Field data and lab data are not the same number

This is the single most common misreading in the topic. A 100/100 Lighthouse score is a lab result. It is genuinely useful for diagnosis — it tells you what to fix and why — but it is not the assessment. Search Console's Core Web Vitals report and the CrUX data behind PageSpeed Insights are the field view, and only pages with enough real traffic have one at all.

INP replaced FID in March 2024

First Input Delay measured only the delay before the browser began handling the first interaction on the page. It was easy to pass, because it ignored how long the handler then took and ignored every interaction after the first.

INP measures the full path — input, processing, and the paint that follows — across all interactions in the visit, and reports a value near the worst of them. Sites that passed FID comfortably can fail INP, and that is not a regression in the site; it is a stricter question being asked. Heavy JavaScript on click handlers, large re-renders and long tasks blocking the main thread are the usual causes.

What it is actually worth for ranking

Google documents page experience as one of many signals, and has said repeatedly that a page with better content will outrank a faster page with worse content. Speed is not a shortcut past relevance.

Where the time usually goes

• Images — uncompressed, oversized for their slot, or served without width and height attributes, which also causes layout shift.
• Third-party scripts — chat widgets, tag managers, analytics stacks and A/B testing tools, each adding main-thread work you did not write.
• Fonts — blocking loads, or swaps that reflow text after paint.
• Late-injected content — cookie banners, promo bars and ads that appear above existing content and push it down; a large share of CLS is this.
• Long tasks — single blocks of JavaScript over 50 ms, which delay every interaction that lands while they run.

None of that requires guessing. PageSpeed Insights names the offending resources, and the Performance panel in Chrome DevTools shows the long tasks directly. The work is unglamorous, and it is legible — you can see the number before and after.