Back to blog

Core Web Vitals in 2026: A Practical Guide to LCP, INP and CLS

Three numbers decide whether users experience your site as fast. Here is what each one measures, why INP is the hardest to fix, and the changes that actually move them.

Performance dashboard showing page load timing metrics

Site speed stopped being a vague feeling a while ago. It has three specific numbers, collected from real Chrome users, and they feed into ranking. In plain language: LCP asks "how long until I saw something?", INP asks "how long until it reacted?", and CLS asks "how much did the page move under my finger?"

The three metrics and their thresholds

  • LCP (Largest Contentful Paint) — when the biggest visible element renders, usually a hero image or headline. Good: under 2.5 seconds.
  • INP (Interaction to Next Paint) — the gap between a tap or click and the first frame showing its result. Good: under 200 milliseconds.
  • CLS (Cumulative Layout Shift) — the sum of unexpected layout movement. Good: under 0.1.

What matters is the 75th percentile of your real users, not one test on your laptop over office fibre. If a quarter of your traffic arrives on a mid-range phone over a congested network, that quarter sets your score.

Field data report showing percentile distribution of page metrics
Field data from real sessions, not a lab run, is what you are judged on.

Why INP is the hard one

INP replaced FID in 2024 and it is far less forgiving. FID only measured the delay before processing started. INP measures the whole round trip: from the tap until pixels change. So if clicking "Add to cart" blocks the main thread for 400ms, your INP is bad even if the page loaded in under a second.

The usual suspects behind a poor INP are always on this list:

  • Heavy event handlers that compute everything inline instead of deferring.
  • Third-party scripts — chat widgets, analytics, ad tags — creating long tasks.
  • Framework re-renders of an entire subtree because one piece of state changed.
  • Interleaved DOM reads and writes causing layout thrashing.

What actually works

For LCP: serve the hero image in AVIF or WebP at the right size, give it fetchpriority="high", and never lazy-load it. Preload critical fonts and use font-display: swap so text paints early. Anything you can do to lower TTFB — server-side caching, edge caching — pulls LCP down with it.

For INP: break long tasks into chunks and yield to the browser between them. Give visual feedback first (disable the button, show a spinner), then do the work. Removing an unnecessary third-party script is usually the single highest-leverage change available, and it requires writing no code at all.

For CLS: set width and height or an aspect ratio on every image, reserve space for banners and ads before they load, and never inject content above existing content after paint.

How to measure

The Chrome UX Report and the Core Web Vitals section of Search Console give you field data, but on a 28-day rolling window. For faster feedback, drop the web-vitals library into your site and send the values to your own analytics. The advantage is attribution: you learn which page and which interaction is responsible, not just that the average is bad.

Then write a performance budget and enforce it in CI — something as simple as "the product page ships no more than 170KB of compressed JavaScript". One rule in a config file prevents dozens of slow regressions that nobody notices until it is too late.

Core Web Vitals in 2026: A Practical Guide to LCP, INP and CLS — 1Core Web Vitals in 2026: A Practical Guide to LCP, INP and CLS — 2Core Web Vitals in 2026: A Practical Guide to LCP, INP and CLS — 3
Written by

Hesam Ziaee, a web developer and programmer who builds fast, user-friendly websites with Vue.js, Quasar and Node.js.