What CrUX measures
The Chrome User Experience Report is a public dataset describing how real people experienced real websites in Chrome. It records loading, interaction and layout stability from visitors who have opted into reporting usage statistics, on sites that are publicly reachable, and publishes the result as an aggregate rather than as individual visits.
It is not a tool you install. If your site is public and receives enough Chrome traffic, it is already in there, and the figures at the top of PageSpeed Insights, the Core Web Vitals report in Search Console and the assessment itself are all reading from it. Data is gathered over a rolling recent window, so the report always describes the recent past rather than today.
Why CrUX matters
It is the measurement Google actually uses. Whatever a synthetic test says, the pass or fail on Core Web Vitals comes from this dataset, which makes it the only speed figure worth reporting to a client as an outcome.
It is also free competitive information. Because the dataset covers any sufficiently visited public site, you can look up a competitor’s domain as easily as your own and see whether the market you are in is fast or slow. That context changes the priority: being slower than everyone in your sector is a different problem from missing an arbitrary target.
Where CrUX goes wrong
Coverage is the first limit. This is Chrome only, so visitors on Safari, on iPhones and in other browsers are absent, and it excludes anyone who has not opted into reporting. A business whose customers are mostly on iPhone is being graded on a fraction of its audience.
Thresholds are the second. A page needs enough visits before it gets its own entry, and below that the tools fall back to the domain as a whole, so a quiet page inherits a verdict earned by the busy ones. Many small sites have no page-level data at all. And because the window rolls, a fix takes weeks to appear — which is where most people lose their nerve and change something else, making the eventual result impossible to attribute.
How to act on it
Start by checking what you are actually looking at. PageSpeed Insights states whether the figures belong to that URL or to the whole domain, and the difference decides whether page-level work will ever show up. Then split phone from desktop, because they usually fail for different reasons and the combined view hides both.
Treat it as a trend rather than a dashboard. Record what you deployed and when, then read the direction over months instead of refreshing it daily; lab tools exist for the day-to-day feedback. Where the domain as a whole is failing, the cause is usually shared — a theme, a tag stack or the hosting — and belongs in a planned page speed project rather than in individual page edits.