How do I read a website speed report?
We recommend reading the report in this order: Device and time → Scores → Loading and interaction metrics → Issue details. Identify the main issues affecting visitors first, then address them one at a time. There is no need to change every setting at once just to chase a perfect score.
1. First, check whether the test is for mobile or desktop, and check the report time
Mobile and desktop tests use different simulated conditions, so differences in scores are normal. Compare them separately. A “Cached results” label means a report from within the last 6 hours is being reused; cached figures will not change immediately after you update your website. When comparing results before and after improvements, use the same URL and device, and make sure the report timestamp has been updated.
2. What do the Performance and SEO scores mean?
Performance:Scores of 90–100 are good, 50–89 need improvement, and 0–49 are poor. This is Lighthouse’s overall score for this simulated test environment, not the speed every visitor will experience. Network conditions, server load, and third-party code can all cause variations, so we recommend reviewing results from multiple tests.
SEO:Checks some basic technical aspects of search engine optimization. A high score does not guarantee high search rankings, nor does it mean that content quality, keyword strategy, or all other SEO work has been addressed.
3. Lighthouse Lab Data: Which Numbers Matter?
LCP (Largest Contentful Paint):How long it takes for the largest main content element on the page to appear—typically a hero image or a large block of text. Lower times are better. If it is slow, start by checking the main image’s file size, image loading order, server response and render-blocking CSS/JavaScript.
CLS (Cumulative Layout Shift):Measures whether text, images, or buttons shift unexpectedly during loading. This value has no time unit; the closer it is to 0, the more stable the layout. Check whether space is reserved for images, fonts change during loading, or content is inserted dynamically.
FCP (First Contentful Paint):The time it takes for visitors to see the first text or image. Lower is better. It does not mean the entire page has finished loading or that buttons respond smoothly.
TBT (Total Blocking Time):The total time the main thread was blocked by long tasks during this simulated page load, expressed in milliseconds. Lower is better. If it is high, start by checking large JavaScript files, plugins, and third-party tracking scripts. TBT is not INP and cannot replace real-user interaction data.
Speed Index:Measures how quickly visible content is progressively displayed, in seconds. Lower is generally better. Useful for comparing results before and after optimization on the same device under similar conditions.
4. Core Web Vitals: Understand real visitors’ experiences
This section uses real-user data collected by Google over the past 28 days, measured at the 75th percentile. For example, an LCP of 2 seconds means that the largest content was displayed within 2 seconds in approximately 75% of the experiences covered by the data. It is not an average across all visitors.
Threshold for “Good”:LCP ≤ 2.5 seconds, CLS ≤ 0.1, INP ≤ 200 milliseconds. INP measures the delay before the screen responds after a click, tap, or keyboard interaction; lower is better. This tool requires complete data for all three metrics to determine whether they all meet the thresholds for a good rating.
If labeled “Entire origin data,” the results represent the overall experience across the same origin and must not be treated as a score for an individual page. Without data, a pass or fail cannot be determined. After website optimization, the accumulated data for 28 days will not immediately reflect the new version in full.
5. Expand the issue details to find the files to check
Click an issue title under “Priority items to review” to expand the file URLs, transfer sizes, estimated savings, or other diagnostic data provided by Google. Clicking a file link opens the resource in a new tab. Share it with your web developer to identify the CSS, JavaScript, or image file involved.
For example, “Reduce unused CSS”:If a file has a transfer size of 104 KiB and estimated savings of 95 KiB, this test has found a large amount of unused content in that file, making it worth checking first. KiB is a unit of data size: 1 KiB = 1,024 bytes. It is neither a duration in seconds nor a performance score.
“Unused” only means that the resource was not used in this particular test scenario. Other pages, mobile layouts, menus, or interactive features may still need it, so do not simply delete the entire file. Estimated savings in milliseconds are not guaranteed to equal the reduction in overall page load time. Different issues may involve the same resources, so the savings cannot simply all be added together. If Google does not provide figures for individual files, this tool will not calculate its own estimates.
6. How should I prioritize improvements?
Start with high-impact items in the details whose purpose you can verify, such as oversized hero images, render-blocking resources, or large amounts of JavaScript. Then adjust caching, compression, and loading methods one at a time. After each change, confirm that the website still functions correctly before comparing a new report for the same device. If the issues mainly involve front-end code or third-party services, upgrading your hosting alone may not resolve them.
Data source:Google PageSpeed Insights Guide. A website speed test is not a security assessment.
Website Speed and SEO Testing FAQ
How is real-user data for Core Web Vitals collected?
Simply enter a URL in this tool to look up the data; you do not need to supply any data yourself. Google anonymously aggregates browsing experiences from Chrome users who meet its data collection criteria into the Chrome UX Report (CrUX). This data reflects real-world user experiences over the most recent 28 days and is updated daily.
Does my website need any software installed or additional configuration?
You do not need to install tracking code on your website or provide access to a Google Analytics or Search Console account to look up publicly available CrUX data from Google. Data may only be available if your website meets Google's inclusion criteria and has enough valid samples.
Why does it say no real-user data is available?
Data may be unavailable if the website is new, has few visitors, or lacks enough samples for the specified page and device. No data returned for this check does not mean the website is slow or has failed Core Web Vitals; any issues with the lookup service need to be checked separately. If the report uses data for the entire origin, this will be clearly indicated and must not be treated as a result for an individual page.
Why does the report still show “Incomplete data” or no INP data when LCP and CLS are available?
Data availability may differ for each metric. Having some values does not mean that data is available for all three metrics. This tool only determines whether all metrics meet the “Good” thresholds when valid LCP, CLS, and INP values are available. If any metric is missing, it displays “Incomplete data.” Missing values are not treated as 0, and this does not mean that the website has failed the assessment. Check whether the report contains data for a single page or the entire site origin, and use the available values as a reference. When INP data is unavailable, Lighthouse's TBT cannot be used as a substitute.
Why do performance scores, SEO scores, or simulated test metrics show “No data”?
“No data” or “—” means the response did not provide a usable value, not a score of 0. A page that fails to load properly, an incomplete test, or a query service issue can all affect the results. If Google provides any notices, read the original messages in the report first. Confirm that the URL is publicly accessible, has no login restrictions, and displays no page errors, then try again later. This tool may reuse cached results for up to 6 hours, so running the test repeatedly may not immediately produce new results. If the issue persists, ask your website developer to review the report. Missing values alone do not indicate poor hosting performance.
What does it mean when Core Web Vitals have “not all met the good thresholds yet”?
This means a complete set of real-user data was retrieved for this check, but at least one metric did not meet the threshold for a good rating: LCP ≤ 2.5 seconds, CLS ≤ 0.1, and INP ≤ 200 milliseconds. This is different from “incomplete data” and does not mean that the entire website or its SEO has failed. Prioritize improvements based on the elevated metrics, and note that the data reflects users’ accumulated experiences over the past 28 days, so changes will not be fully reflected immediately. If origin-level data for the entire website is used, the results cannot be attributed solely to the current page.
If no file details or priority improvements are listed, does that mean the website has no issues?
Not necessarily. Some checks provide an overall diagnosis without listing individual files; this tool also displays only the items available in the current response. A lack of details does not mean there are no issues, nor does it mean that all SEO or security checks have been completed. Interpret the results alongside the individual metrics, Google's notices, and how your website is actually used. If code changes are needed, ask your website designer or developer to investigate further.
Will clicking “Start Test” a few more times make data available?
No. Running the test again does not add real visitor samples. This tool also reuses cached results for the same URL and device for 6 hours. Real-user data must accumulate gradually through actual everyday browsing.
Is the Performance score still useful without real-user data?
Yes. Lighthouse simulated tests can still help identify ways to improve images, code, and loading speed, but they cannot replace real visitor experience data. TBT and INP are different metrics; TBT cannot be used to determine whether real users' INP meets the passing threshold.
Learn more:Google CrUX Data Collection Requirements、About PageSpeed Insights。