Domains, CDN & Web Delivery

Core Web Vitals

Definition

The three cover loading, responsiveness and visual stability in turn. Largest Contentful Paint records when the biggest visible element has finished rendering — good is 2.5 seconds or under, needs improvement runs to 4 seconds, and anything slower is poor. Interaction to Next Paint measures the worst delay between a tap or click and the screen updating in response — good is 200 ms or under, needs improvement runs to 500 ms, poor is beyond that. Cumulative Layout Shift scores how much content jumps around while loading, on an unbounded scale where 0.1 or less is good and above 0.25 is poor. INP replaced First Input Delay in March 2024. A page passes only if all three land in the good band at the 75th percentile of visits over a rolling 28 days, counted separately for mobile and desktop.

Why It Matters

Because the assessment is at the 75th percentile of real traffic, a site can feel fine on a developer's laptop and still fail — the quarter of visitors on older phones and slower networks are the ones being measured. The ranking effect is genuinely modest, a tie-breaker rather than a lever, but the underlying behaviour is not: a layout that shifts under a reader's thumb makes them tap the wrong link, and a 3 second paint loses a share of visitors before anything appears. Static pages tend to start ahead, because there is no server rendering time and the files can sit on an edge node near the visitor. The common failure is not the hosting at all — it is an oversized hero image, a blocking font, or an advert slot with no reserved height.

How It Works

Two sorts of data exist and they answer different questions. Field data comes from the Chrome User Experience Report, collected from real Chrome users who opted in, and is what Search Console and the ranking signal use; it is a 28-day rolling window, so a fix takes weeks to show. Lab data comes from a tool such as Lighthouse running a single simulated load, which is reproducible and useful for debugging but cannot produce a real INP figure, since nobody clicked anything. In the browser, each metric arrives through a PerformanceObserver — largest-contentful-paint, event and layout-shift entry types — and the web-vitals JavaScript library wraps those into one number per metric. Of the three, Largest Contentful Paint is usually where the work is, and it is the one most directly affected by how the files are delivered.

Real-World Example

A recruiter publishes a hiring page on 99helpers at join-halden.99helpers.site with a 2.4 MB PNG banner. Field data shows LCP at 4.1 seconds on mobile, CLS at 0.28 because a web font swaps in late and reflows the headline, and INP comfortably inside 200 ms since the page barely runs any script. Exporting the banner as a 180 KB WebP and giving the headline element a fixed height brings LCP to 1.6 seconds and CLS to 0.02 — two failures cleared without touching the copy.

Common Mistakes

  • Judging a site by its Lighthouse score alone — that is one simulated load, while the ranking signal uses 28 days of real visits
  • Fixing a problem and expecting the numbers to move the same day, when the field window takes weeks to catch up
  • Optimising the desktop experience while the failing bucket is mobile, which is assessed separately
  • Treating a passing average as a pass — the threshold is the 75th percentile, so a good median can still hide a failing quarter of visits

Related Terms

Put a file online in seconds

Drop in a document, an image, a page or a whole static website and share the link — free, with no build step and no server to set up.

Host a file free →