Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Measure Largest Contentful Paint (LCP), Interaction to Next Paint (INP), and Cumulative Layout Shift (CLS) first: they cover loading, responsiveness, and visual stability. Add First Contentful Paint (FCP) and Time to First Byte (TTFB) to investigate loading delays, and Total Blocking Time (TBT) as a lab-only diagnostic clue for interaction problems. Use controlled lab tests to reproduce regressions and real-user field data to learn what visitors actually experience; judge Core Web Vitals at the 75th percentile, not by a single score or average.
Which web performance metrics should you measure?
Use Core Web Vitals as the primary outcome measures. The current set consists of LCP, INP, and CLS. Each corresponds to a distinct part of the experience, so passing one does not compensate for failing another. The thresholds below are Chrome guidance categories, not results from a dated performance study.
| Metric | Experience measured | Good | Needs improvement | Poor | How to use it |
|---|---|---|---|---|---|
| LCP | When the page’s likely main content appears | At or below 2,500 ms | Above 2,500 ms through 4,000 ms | Above 4,000 ms | Core Web Vital; available in field and lab measurements |
| INP | Responsiveness across interactions during a page visit | At or below 200 ms | Above 200 ms through 500 ms | Above 500 ms | Core Web Vital; meaningful measurement requires interaction and is best assessed with field data |
| CLS | Unexpected visual movement | At or below 0.10 | Above 0.10 through 0.25 | Above 0.25 | Core Web Vital; measurable in field and lab, though a load-only lab run can miss later shifts |
| FCP | Time until the first foreground content appears | At or below 1.8 s in PageSpeed Insights guidance | Not stated | Not stated | Supporting loading diagnostic, not a Core Web Vital |
| TTFB | Time until the browser receives the first byte of the response | At or below 0.8 s in PageSpeed Insights guidance | Not stated | Not stated | Supporting loading diagnostic; PageSpeed Insights labels it experimental |
| TBT | Main-thread blocking during page load | No Core Web Vital threshold in this guidance | No Core Web Vital threshold in this guidance | No Core Web Vital threshold in this guidance | Lab diagnostic that can signal potential INP problems; it is not INP |
The threshold categories and definitions are from Chrome for Developers’ PageSpeed Insights guidance. FCP, TTFB, and TBT help explain what may be contributing to a poor experience; they do not replace the three Core Web Vitals.
Why INP needs more than a page-load test
INP reflects responsiveness across interactions during a visit. A test that merely loads a URL without clicking, typing, or otherwise interacting cannot directly establish how responsive the page is. Field data can include the interactions visitors actually perform; lab testing can still help reproduce and diagnose a particular interaction when the test exercises it.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minute#1 Best Overall
Why a clean-looking load may still have poor CLS
CLS concerns unexpected movement, which may occur after initial rendering—for example, later in a visit. A lab run that stops observing before that event can miss it. Compare the test’s observation window and interactions with the conditions under which the shift appears for users.
How should you interpret the numbers?
Use the 75th percentile and inspect the spread
For each Core Web Vital, the recommended assessment is that at least 75% of page visits meet its good threshold. Look at the 75th percentile and the distribution behind it: a median or overall average can conceal a substantial slow tail. These are Chrome and web.dev measurement recommendations, not a claim that every visitor or every page load must have an identical result. See web.dev’s guide to measuring Web Vitals.
Rank #2
Break results out by page group and, where available, by user conditions such as device or connection. Site-wide aggregates can hide a template-specific problem, while a single unusually slow visit does not establish that all pages are failing. Compare like with like: use the same metric, percentile, time window, and sufficiently similar population.
Field and lab measurements answer different questions
- Field data records real users under varied devices, networks, locations, cache states, content, and interactions. The Chrome UX Report (CrUX) provides aggregated real-user experience data; a site’s own real-user monitoring (RUM) can provide more detailed, timely page-view telemetry.
- Lab data comes from a controlled test such as Lighthouse. It is useful for repeatable diagnostics and catching regressions before release, but it cannot reproduce every real visitor’s conditions or behavior.
A field score and a Lighthouse result can differ without either being erroneous: the populations and conditions differ. As the Chrome team puts it, “Only field measurement can accurately capture the complete picture.” (Web Vitals.)
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWhich tools should you use for each job?
Quick check for one page
Enter the URL in PageSpeed Insights. When available, it presents CrUX field data alongside Lighthouse lab diagnostics. A page may not have enough field data to show a metric, so the lab report is not a substitute for missing real-user evidence.
Find patterns across a site
Use the Core Web Vitals report in Google Search Console to find groups of similar URLs with issues. It is intended for site-wide patterns, not as the best status lookup for one URL; use a page-level test for that. Search Console describes a 28-day validation session after a fix, so validation is a monitoring window rather than an instant retest.
Rank #4
Debug locally or reproduce a condition
- Use Chrome DevTools’ Performance panel to inspect local Core Web Vitals and the events around a problem.
- Run Lighthouse in DevTools, as a package, or in continuous integration to check controlled changes and catch regressions.
- Use WebPageTest when you need to specify device or network conditions for a test.
These tools and their field-versus-lab limitations are described in web.dev’s measuring guide.
Monitor actual visitors with RUM
For detailed, timely measurements by page view, supplement CrUX with your own RUM. The web-vitals JavaScript library is one implementation option; sending its measurements to an analytics or reporting endpoint is necessary to make them useful. Choose reporting dimensions that help distinguish page groups and user conditions without treating small segments as representative when their data is sparse. The measurement guide documents both RUM and CrUX’s calendar-month reporting breakdown.
Best Value
- Used Book in Good Condition
A practical measurement workflow
- Choose the question. For a single URL, start with PageSpeed Insights or a controlled browser test. For recurring site-wide issues, inspect Search Console groupings. For detailed production behavior, instrument RUM.
- Record all three Core Web Vitals. Capture LCP, INP, and CLS, plus the 75th-percentile result and distribution where field data is available. Note when a metric has no field data rather than treating a lab value as equivalent.
- Add diagnostic metrics when they help explain a failure. Use FCP and TTFB to investigate loading delays; use TBT to look for main-thread blocking that may relate to interaction responsiveness.
- Make the conditions explicit. Record whether the result is field or lab, the page or URL group, the test setup, and relevant device, network, location, cache, content, and interaction conditions when known.
- Reproduce and investigate. Use DevTools or a repeatable Lighthouse/WebPageTest run to inspect the relevant loading, interaction, or layout behavior. Do not infer the cause from a score alone.
- Validate after the change. Repeat the controlled test under comparable conditions, then watch field data. Search Console’s described validation process takes 28 days; it is not an immediate confirmation.
Common measurement mistakes and how to avoid them
- Treating a Lighthouse score as the whole user experience: lab conditions do not represent every device, connection, location, or interaction. Pair repeatable diagnostics with field evidence.
- Calling TBT “INP”: TBT is a lab metric based on a different calculation. Use it as a clue about possible main-thread causes, not as an INP result.
- Using only a median or average: these can hide slow visits. Review the 75th percentile and distribution for Core Web Vitals.
- Assuming a page-load run covers a full visit: it may not trigger interactions needed for INP or observe shifts that happen later and affect CLS.
- Attributing a field-status change only to a code release: traffic mix, network conditions, browser changes, or upstream service latency can also move results. Check whether the affected population or conditions changed before assigning a cause.
Or skip the browser setup
For screenshot capture in an automated workflow, ScreenshotNeo offers a one-request API. A screenshot is useful for visual inspection, but it does not measure LCP, INP, or CLS and should not replace performance measurement.
cURL example, using the documented API call and substituting the page you want to capture:
Quick Recap
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpSee the ScreenshotNeo documentation for API options. It removes cookie banners, newsletter popups, and chat widgets before capture; bot checks, blank pages, and failed loads are not billed. Its MCP server lets AI agents take screenshots. The free plan includes 1,000 screenshots a month with no card, and paid plans start at $5 for 3,000. Sign up for the free plan.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
