Improving Core Web Vitals

The three numbers, and what they mean

Three metrics matter, each with a threshold below which Google considers the experience good:

• LCP (Largest Contentful Paint) — when the largest element is on screen. Good: under 2.5 seconds. Usually your hero image or the first heading.
• INP (Interaction to Next Paint) — how quickly the page responds to a click or tap. Good: under 200 milliseconds. This replaced the older FID metric in 2024.
• CLS (Cumulative Layout Shift) — how much the layout jumps while loading. Good: under 0.1. This is the metric behind the feeling of pressing the wrong button.

Lab data and field data are not the same thing

The audit measures two things. First a lab measurement: we load your page on an emulated phone over a Slow 4G connection and measure LCP, CLS and Total Blocking Time. That measurement is reproducible and available immediately, including for a site with no visitors.

Second, we pull field data from the Chrome UX Report: LCP, INP and CLS as real visitors experienced them over the past 28 days. That is what Google itself looks at. The catch is that this data only exists if your domain has enough traffic — if it does not, the report says there is no field data rather than pretending otherwise.

When the two diverge sharply, that is information in itself. Fast in the lab but slow in the field usually means your real visitors are on slower devices than you test with, or come from a region further from your server.

What actually makes a page slow

The causes are rarely exotic. The same four turn up in virtually every report, and the audit points at them concretely with filename and size:

• Images served far larger than they are displayed, or still in JPEG where WebP or AVIF would cost half
• JavaScript bundles that block rendering because they sit in the head without defer or async
• Static files without effective caching, so returning visitors download everything again
• Third-party origins — fonts, analytics, chat widgets — with no preconnect configured, so every connection starts late

Start with CLS, it is the cheapest

Shifting layout is almost always fixed by setting width and height on your images, and reserving space for elements that load later such as banners and ads. That is an hour's work and the difference is immediately visible.

LCP takes more effort but returns the most. Usually the win comes from a single image: your hero. Serve it in a modern format, at the right dimensions, with a preload hint, and your LCP often drops under the threshold in one move.

INP is the hardest, because it is about how much JavaScript occupies your main thread. There is rarely a quick trick for it — that is cleanup, or code that loads only when it is needed.

Frequently asked questions

How much do Core Web Vitals affect my ranking?

They are one signal among many and weigh less than the relevance of your content. Between two comparable pages they can decide it; a slow page with a far better answer still beats a fast empty one.

Why does my score differ between measurements?

Lab measurements are sensitive to network and server noise. Look at the direction across several runs, not at one isolated result. Field data from the Chrome UX Report is steadier because it is averaged over 28 days.

Why is there no INP in my report?

INP can only be measured from real interaction by real visitors. If your domain has too little traffic for the Chrome UX Report, we show Total Blocking Time from the lab measurement instead — a good predictor, but not the same number.