Chrome UX Report (CrUX) is Google’s aggregated field dataset of eligible real Chrome user experiences, available for popular public URLs and origins.
The Chrome UX Report, usually called CrUX, is Google’s dataset of aggregated real-user experience data from eligible Chrome browsing. It reports field data: observations from users in the wild, not a simulated page load. CrUX can therefore show how a population experienced a public page or origin, while a laboratory test helps diagnose a controlled visit.
What CrUX measures
CrUX publishes distributions and percentiles for user-experience metrics, including Core Web Vitals. The dataset is privacy-preserving and aggregated; it is not session replay, analytics, a synthetic score, or a complete census of every visitor. Its usefulness is that it reflects varied devices, networks and browsing conditions that a single test machine cannot reproduce.
Field data is descriptive, not a diagnosis. A poor LCP distribution can tell you that users are experiencing slow largest-content rendering, but it does not identify the responsible asset, cache rule or backend dependency. Pair CrUX with page inspection, network evidence and a repeatable lab run before changing implementation.
URL, origin and eligibility
A URL record is specific to one normalized page address. An origin record combines eligible page-load experiences across the origin. That distinction matters: an origin can look healthy while a revenue-critical or content-critical URL is poor, and a URL with low traffic may have no record even though its origin does.
| Question | Use | Limit |
|---|---|---|
| How do users experience this page? | Query a URL record | It may be unavailable when the page lacks enough eligible traffic. |
| How does the site perform overall? | Query an origin record | It can hide differences between templates or individual pages. |
| What caused a slow load? | Run a controlled lab test and inspect the page | A lab test does not measure the same user population. |
Chrome documents eligibility separately for users, origins and pages. Public discoverability and sufficient popularity are required, and the precise popularity threshold is not published. A missing URL record is therefore an availability state, not evidence that the page is fast, slow, unimportant, or excluded from search for one specific reason.
Read field data carefully
- Record the exact identifier: URL or origin, protocol and host included.
- Choose the form factor deliberately, or state that the result is aggregated across devices.
- Note the collection period and whether the response contains field data, a partial metric set, or no record.
- Read the distribution and the relevant percentile together; a single summary value cannot describe every user experience.
- Compare like with like over time. Do not combine field data, analytics sessions and synthetic scores into one invented metric.
I label incomplete field data as incomplete. An unavailable measurement is not a green result, and an origin-level result is not proof about every template. That simple separation prevents a common reporting error: turning a missing URL-level record into a performance conclusion.
Investigate and verify
When CrUX indicates a problem, start with the affected scope. Check whether the issue appears on URL and origin, then reproduce an appropriate page in a controlled environment. Inspect the rendered HTML, response timing, critical resources and the largest visible element. A change should be reversible and measured again only after enough field-data time has elapsed for the dataset to reflect new visits.
- Verify the page returns a public, successful response after redirects; eligibility depends on discoverability.
- Use a URL-level record for a page decision whenever it exists, and describe an origin result as an aggregate.
- Separate a field trend from causation: a change in CrUX may coincide with releases, traffic mix or devices without proving why it happened.
- Keep the raw response scope and date with the report so a later comparison uses the same identifier and dimensions.