{"canonical_url":"https://engineer.company/glossary/core-web-vitals/","content_markdown":"# Core Web Vitals\n\n## What it is\n\nGoogle's web.dev documentation picks out three metrics from its wider Web Vitals\nprogramme and calls them core: Largest Contentful Paint for loading, Interaction\nto Next Paint for responsiveness and Cumulative Layout Shift for visual\nstability. Each describes something a visitor notices, applies to every page and\ncan be measured in the field, in real visitors' browsers.\n\nLargest Contentful Paint (LCP) is the time from the start of navigation to the\nmoment the largest image, text block or video visible in the viewport has\nrendered. It is good at 2.5 seconds or less and poor above 4 seconds.\n\nInteraction to Next Paint (INP) watches every click, tap and key press during a\nvisit and takes the whole wait the visitor sees: the delay before the handlers\nstart, the time they run and the time until the next frame is painted. The\npage's INP is the slowest of those, with one outlier dropped for every 50\ninteractions. Scrolling, hovering and zooming are left out. It is good at 200\nmilliseconds or less and poor above 500. INP replaced First Input Delay on 12\nMarch 2024, having been introduced as an experimental metric in May 2022; First\nInput Delay measured only the delay before the first interaction's handlers\nbegan.\n\nCumulative Layout Shift (CLS) scores visible elements that move without being\nasked to. Each shift scores the share of the viewport the moving elements cover\nmultiplied by how far they move. Shifts less than a second apart, up to five\nseconds in all, form a session window, and CLS is the total of the worst window.\nShifts within 500 milliseconds of a tap, click or key press are excused;\nscrolling and dragging excuse nothing. It is good at 0.1 or less and poor above\n0.25.\n\nThe verdict is read at the 75th percentile of page loads, separately for mobile\nand desktop, and a page passes only when all three are good there. A \"good\" line\nwas accepted only if at least a tenth of origins already met it.\n\nUnderneath are browser interfaces with their own specifications, the Largest\nContentful Paint, Layout Instability and Event Timing APIs, and a few lines show\nthe raw material of CLS on any page:\n\n```js\nnew PerformanceObserver((list) =\u003e {\n    for (const shift of list.getEntries()) {\n        if (!shift.hadRecentInput) console.log(shift.value, shift.sources);\n    }\n}).observe({ type: \"layout-shift\", buffered: true });\n```\n\nThat logs individual shifts, not the metric; for the three as Google's tools\nreport them, web.dev points to the `web-vitals` library.\n\nThe term is easy to confuse with the Lighthouse performance score, which comes\nfrom one simulated load in a lab, with no visitor in it; the Core Web Vitals\nassessment describes real visits at the 75th percentile, and a page can do well\non one and badly on the other.\n\n## Why it matters\n\nThe definitions leave open where the numbers come from, and that is the decision\na site has to make: steer by lab estimates, or pay for field data.\n\nField data is what the thresholds are written for. The\n[Chrome User Experience Report](https://developer.chrome.com/docs/crux) collects\nanonymised measurements from real Chrome users and feeds\nPageSpeed Insights and Search Console's Core Web Vitals report without the site\nadding anything, but only once a page or origin has enough visitors. The\nalternative is a script in the visitor's browser that sends every reading to a\nserver: a privacy decision, and a cost carried on every page view.\n\nLab data costs neither. Lighthouse loads the page on a simulated device and\nnetwork, before a change ships and as often as wanted. What it measures is a\nvisit nobody made. It sees LCP and CLS during the load, but with nobody pressing\nanything it has no INP; web.dev offers Total Blocking Time as the closest lab\nmeasure and says it is no substitute.\n\nThe workable split is both, held apart: lab numbers as budgets that catch a\nchange for the worse, field numbers, where they exist, as the only verdict on\nwhat visitors get.\n\n## What it looks like in practice\n\nA planning record shows a Core Web Vitals job was considered and declined: the Chrome UX\nReport needs an API key and enough traffic to report, and the pages are already held to\nsize budgets in the quality gate. The site gathers no browser-side vitals itself: its\ncredits page says it [ships no executable JavaScript or browser analytics\ncode](/credits/). There is no site-supplied script to hold up a click or key press,\nalthough layout and rendering can still delay interaction. The same page points readers\nto PageSpeed Insights, which shows measurements from real visits once a page has enough\nof them.\n\nWhat the site measures instead is the lab. Its Lighthouse configuration loads a\nfixed list of pages three times each under simulated mobile throttling and\nreports a failure when one of the English or Danish pages crosses any of these\nlines: LCP over 4,200 milliseconds, First Contentful Paint over 1,600, CLS over\n0.02 and Total Blocking Time over 100, with slightly wider paint limits on the\nUkrainian pages. These are budgets, not the field thresholds, written after the\nrun-to-run spread had been measured at 299 milliseconds at worst and 80 at best,\nbecause a line inside that spread passes or fails by chance. The project's notes\nput most of the LCP down to render delay behind one render-blocking stylesheet.\nWeight that arrives before any content had already been cut:\n[the stylesheet bundle weighed 87 KB with a base64 font inside it and 31 without](/portfolio/cut-the-stylesheet-bundle-from-87-kb-to-31-kb/).\nThe runs take minutes, so they happen on demand, outside the gate.\n\nThe CLS budget, five times stricter than the 0.1 line, caught a case worth\nkeeping. One category page failed at 0.107. The owner opened it, saw nothing\nwrong and said so. Lighthouse had named the parallax gears as the culprit, and\nthe first written diagnosis believed it. A controlled test with a real wheel\ngesture scored 0.2477 whether the gears moved or not, while the heading and the\nlast paragraph moved by zero pixels throughout. The cause was the footer: it is\nfixed to the viewport and changes height as its colophon opens, and the metric\nscores that box against the viewport although a fixed element cannot push\nanything. The budget stayed where it was; if the number is ever silenced, the\nplan is an exception for that one address.\n\n## Where it goes wrong\n\n**A lab report leaves INP out and the page is assumed responsive.** Lighthouse\nshows LCP and CLS, Total Blocking Time is low, and slow clicks in production\ncome as a surprise. A simulated load has nobody interacting, and Total Blocking\nTime only counts main-thread blocking during that load. Read it as a warning\nlight, time real interactions by hand in the browser's performance tools, and\njudge INP from field data where there is any.\n\n**The element a tool blames turns out to be innocent.** The report names a node\nfor a layout shift, changing that node moves the score by nothing, and the text\non screen stays put. Attribution names nodes, not causes; the shift may belong\nto a fixed overlay that changes height, which counts as movement and is not\nexcused by scrolling. Switch suspects off one at a time, measure whether content\nmoves, and animate size with `transform`, which web.dev recommends because it\ncauses no layout shift.\n\n**Content jumps down as images and late arrivals load.** Readers lose their\nplace and field CLS climbs past 0.1. web.dev lists the usual causes: images and\nvideo of unknown size, web fonts that render larger or smaller than the\nfallback, and widgets that resize themselves. Give every image `width` and\n`height` attributes so the browser reserves its box, as this site's image\npartials do, and reserve space for anything inserted later.\n\n**A healthy average hides a failing assessment.** A dashboard shows a mean LCP\nwell under 2.5 seconds, yet the page is rated as needing improvement. The\nassessment uses the 75th percentile, per device class, and a slow quarter of\nmobile visits disappears inside a mean that includes fast desktops. Report and\ntarget the 75th percentile, split by device.\n\n## Further reading\n\n- [Web Vitals](https://web.dev/articles/vitals) — the three metrics, their\n  thresholds and the tools that report them.\n- [Defining the Core Web Vitals metrics thresholds](https://web.dev/articles/defining-core-web-vitals-thresholds)\n  — why the 75th percentile, and how each line was tested.\n- [Interaction to Next Paint](https://web.dev/articles/inp) — what counts as an\n  interaction and how INP differs from First Input Delay.\n- [Largest Contentful Paint API](https://wicg.github.io/largest-contentful-paint/)\n  — which elements are candidates and when their time is taken.\n- [Layout Instability API](https://wicg.github.io/layout-instability/) — the\n  layout shift score, recent input and the sources of each shift.\n- [Event Timing API](https://w3c.github.io/event-timing/) — how interaction\n  latency is exposed to scripts.\n\n\u003chttps://engineer.company/glossary/core-web-vitals/\u003e\n","date_modified":"2026-10-30T12:43:03Z","description":"What the three Core Web Vitals measure, why their pass marks come from real visits, and how a lab budget can flag a shift that no reader sees.","formats":{"gemtext":"https://engineer.company/glossary/core-web-vitals/index.gmi","gopher":"https://engineer.company/glossary/core-web-vitals/gophermap","html":"https://engineer.company/glossary/core-web-vitals/","json":"https://engineer.company/glossary/core-web-vitals/index.json","markdown":"https://engineer.company/glossary/core-web-vitals/index.md","pdf":"https://engineer.company/glossary/core-web-vitals/index.pdf"},"kind":"page","language":"en","title":"What is Core Web Vitals?","url":"https://engineer.company/glossary/core-web-vitals/","version":1}