Using the Chrome UX Report on PageSpeed Insights
PageSpeed Insights (PSI) is a tool for web developers to understand what a page's performance is and how to improve it. It uses Lighthouse to audit the page and identify opportunities to improve performance. It also integrates with the Chrome UX Report (CrUX) to show how real users experience the page and the origin in aggregate. In this guide, learn how to use PSI to extract insights from CrUX and better understand the user experience.
Reading the data
To get started, go to https://pagespeed.web.dev/ and enter the URL of the page you want to test and click Analyze.
After a few seconds, Lighthouse audits will be performed and you will see sections with data from CrUX ("Discover what your real users are experiencing") and Lighthouse ("Diagnose performance issues"). CrUX is a collection of real-user experiences from the field, while Lighthouse is a controlled test in the lab.
In the real-user experiences section, metrics are grouped by Core Web Vitals and other notable metrics. The three Core Web Vitals metrics are listed first: Largest Contentful Paint (LCP), First Input Delay (FID), and Cumulative Layout Shift. These are followed by the other notable metrics: First Contentful Paint (FCP), Interaction to Next Paint (INP), and Time to First Byte (TTFB).
Core Web Vitals
- LCP measures the time until the page displays what is likely its main content, like a hero image or heading.
- FID measures the interactivity of the page, from the user's first interaction to the time the page responds to it.
- CLS measures the degree of layout instability on the page, due to shifts like asynchronously loaded content being injected.
Other notable metrics
- INP is an pending Core Web Vital metric that measures the latency of all interactions a user has made with the page, and reports a single value which all (or nearly all) interactions were below.
- FCP measures the time until the page displays something in the foreground, like some text or a logo.
- TTFB is an experimental metric that measures the time between the request for a resource and when the first byte of a response begins to arrive.
This table describes how values for these metrics are categorized as either "good", "needing improvement", or "poor".
The Core Web Vitals metrics are the most important metrics, and are directly experienced by real users. The other metrics may be used as diagnostics, but they're not necessarily perceptible by real users. Thus, these metrics are not factored into the Core Web Vitals assessment, and failing to meet the "good" thresholds only matters if they're adversely affecting the Core Web Vitals.
There are three ways that the user experience is displayed in PSI:
- a label summarizing the page as passing or not passing the Core Web Vitals assessment
- percentiles measured in seconds or milliseconds (CLS is unitless)
- a distribution representing the percent of "good", "needing improvement", and "poor" experiences is available by clicking on Expand view in the top right of this section.
In the screenshot above, the page is labelled as "passing" the Core Web Vitals assessment. To pass, the percentile must be categorized as "good" in all three Core Web Vitals. Otherwise, the assessment appears as "failed". Some pages may not have sufficient FID data, in which case the page is assessed on the other two Core Web Vitals metrics.
The percentiles shown for all metrics correspond to the 75th percentile. In statistics, a percentile is a measure that indicates the value below which a given percentage of samples fall. For example, the screenshot above shows that FID's 75th percentile is 13ms, meaning that 75% of FID experiences are faster than 13ms. These values are color-coded according to the threshold table above where "good" values are green, values "needing improvement" are orange, and "poor" values are red.
Finally, the distributions for each metric are illustrated using the "good", "needs improvement", and "poor" grouping. For example, LCP experiences on this page are "good" (less than 2.5 seconds) 90% of the time. FID is "poor" (at least 300 milliseconds) 1% of the time. These distributions represent all user experiences on the page and their shapes indicate the tendency to be either "good" or "poor".
Summary of origin performance
PSI also includes a summary of origin performance. This is an aggregation of user experiences across all pages of an origin. You can get the same stats for an entire origin that are available for individual pages. This data is closely aligned with what is available on BigQuery, while the page-level performance is not made available to query.
There is one major difference between the origin-level data on PSI versus BigQuery. The datasets on BigQuery are released once a month and encompass the data from the previous calendar month. For example, the 202005 dataset includes all user experiences that occurred in May 2020. On the other hand, PSI aggregates new data every day encompassing the previous 28 days. So the results you see today may be different tomorrow and they would not necessarily be the same as what you'd see in the current month's aggregation on BigQuery.
Responses when URL data is not available in CrUX
If the URL you entered is not available in CrUX, PageSpeed Insights will attempt to fallback to origin-level data as shown below. Clicking on the icon beside the disabled This URL button will show more explanation.
If the origin-level data is also not available in CrUX, then PSI is not able to display this section and you will see No Data. Lab data from Lighthouse is still available to give you an approximation of the page's performance.
When would I use PageSpeed Insights as opposed to other tools?
PSI combines the real-world user experience data of CrUX with the lab-based performance diagnostics of Lighthouse. This makes it easy to see how fast a page is experienced and how to make it faster all in one place. The daily aggregation of field data in PSI makes it a great place to more closely monitor origin or URL performance than tools with less frequent aggregations.
Are there any limitations to using PageSpeed Insights?
PSI only provides the most recent daily aggregation, so you wouldn't necessarily be able to see how a site's performance is trending. There are also some non-vital metrics included in the CrUX dataset that are not exposed in PSI.
Where can I learn more about PageSpeed Insights?
Check out the PSI documentation for more info.