Three metrics Google uses to score how your site actually feels. CorbaCount captures them straight from your visitors' browsers - no synthetic test, no separate tool.
Each one measures a different way a page can feel broken: slow to show up, slow to respond, or jumping around while you read.
How long until the biggest visible thing on the page (usually the hero image or headline) finishes loading. Good is under 2.5 seconds, per Google's documentation. Above 4 seconds is "poor" in Google's scoring.
The lag between a tap or click and the page actually responding to it. Replaced FID in March 2024 because it captures the full interaction, not just the first. Good is under 200ms, per Google's documentation.
How much content jumps around as the page loads (think: a button you were about to tap moving at the last second). Good is under 0.1, per Google's documentation. Mostly caused by images without dimensions and late-loading ads or banners.
Field data from the same visitors that count towards your traffic - not a lab test on a synthetic machine.
The same tracking snippet that counts the pageview starts listening. Nothing extra to install.
The tracker hooks into PerformanceObserver in real visitor browsers - the same source Chrome's CrUX report uses.
Beaconed back when the visitor navigates away, so the measurements are complete and never block your page.
Averaged per day, per page and per device class, so the report stays fast however much traffic you get.
Google assesses mobile and desktop on their own, so the report does too. Mobile is usually where the bad surprises live: a hero image that loads in a second on your laptop can take three on a mid-range phone.
| Device | LCP | INP | CLS | Samples |
|---|---|---|---|---|
| Mobile | 2.71s | 236ms | 0.09 | 21.4K |
| Desktop | 1.20s | 88ms | 0.04 | 15.1K |
| Tablet | 1.99s | 142ms | 0.06 | 1.9K |
A good site-wide average can hide one slow template. The per-page table shows LCP, INP and CLS for your 20 most visited pages, so you know exactly where to spend your engineering time.
| # | Path | LCP | INP | CLS | Samples |
|---|---|---|---|---|---|
| 1 | / | 1.62s | 124ms | 0.04 | 9.8K |
| 2 | /blog/how-to-track-ai-traffic | 2.14s | 182ms | 0.07 | 6.2K |
| 3 | /pricing | 1.38s | 96ms | 0.02 | 4.9K |
| 4 | /features | 2.89s | 214ms | 0.13 | 3.1K |
| 5 | /docs/install | 1.19s | 74ms | 0.01 | 2.7K |
| 6 | /blog/core-web-vitals-checklist | 4.31s | 268ms | 0.28 | 1.8K |
| 7 | /checkout | 2.46s | 540ms | 0.05 | 1.3K |
| 8 | /contact | 1.31s | 102ms | 0.03 | 914 |
Google has used Core Web Vitals as a ranking signal for desktop and mobile search since 2021. Poor vitals don't tank your rankings overnight, but everything else being equal, the faster site wins.
Slow pages get abandoned. Bounce rate climbs sharply once LCP crosses 3 seconds. Layout shifts make people misclick - and lose trust in your interface.
Lighthouse runs on a fixed simulated device. Your visitors run on five-year-old Android phones on bad Wi-Fi. Only field data tells you what they actually see.
The Chrome User Experience Report uses the exact same browser APIs CorbaCount taps into. If your dashboard shows a bad number, Google sees it too.
Lighthouse and PageSpeed Insights run your page once, on one simulated device, on one throttled connection. Useful for debugging - but it is not what your visitors experience.
Lighthouse, PageSpeed Insights
CorbaCount, Chrome UX Report
Because the spread between visitors is wide, Google does not look at your average - it looks at the 75th percentile. Your LCP only counts as "good" if at least 75% of real visits load in under 2.5 seconds. CorbaCount measures the vitals inside the same tracking snippet that counts your visitors, using the browser's native PerformanceObserver API on every real pageview. There is nothing extra to install beyond the one-line snippet. Run Lighthouse to find why a page is slow; use field data to know whether it is slow. See how the report fits in the rest of the dashboard in the dashboard tour.
Because slow, janky pages make search results look bad. Google folded the vitals into its page experience signals in 2021: between two pages with similar content, the one that loads faster and shifts less tends to rank higher. It is a tiebreaker, not a magic lever.
At the 75th percentile of real visits: LCP at or under 2.5 seconds, INP at or under 200 milliseconds, CLS at or under 0.1. "Poor" starts at 4 seconds, 500 milliseconds and 0.25 respectively. Anything between is "needs improvement". Mobile and desktop are assessed separately.
Field data, because it is what Google's ranking signal actually uses and what your visitors actually feel. Lab tools are reproducible debuggers: great for before/after comparisons while you fix something, blind to real devices and networks. Trust the field number, debug with the lab tool.
Yes - conversions. Visitors abandon pages that feel slow long before any search-ranking effect shows up, and layout shifts cause misclicks that erode trust. Vitals sit next to your traffic and the rest of your domain health signals, so you see speed and outcomes in one place.
Core Web Vitals are included on every CorbaCount plan, no upgrade needed. Drop your email and we will let you know when it is your turn.
