Iron Goo
Iron Goo robot waits for a phone to load while sitting on a city bus at dusk

Core Web Vitals and Mobile Speed Reference Card

Atamyrat Hangeldiyev
Atamyrat Hangeldiyev
Systems Architect
March 11, 2026
On this page
UX

Only 48% of mobile websites passed all three Core Web Vitals in the Web Almanac 2025 performance chapter, compared with 56% on desktop. The figures describe websites at the origin level using Chrome UX Report data from July 2025; the chapter was published on 15 January 2026.

Core Web Vitals are three field metrics: Largest Contentful Paint (LCP) for loading, Interaction to Next Paint (INP) for responsiveness, and Cumulative Layout Shift (CLS) for visual stability. Each is judged at the 75th percentile of real page loads, with mobile and desktop measured separately. A fast developer laptop can make a slow customer experience easy to miss.

Speed is one part of user experience for a small business, alongside whether customers can understand and complete the journey. Sources read on 29 October 2026.

MetricGoodNeeds improvementPoorWhat it measuresWhere to see it
LCP2.5 s or lessOver 2.5 s to 4.0 sOver 4.0 sTime until the largest eligible visible content is renderedPageSpeed Insights field data and Lighthouse; Search Console
INP200 ms or lessOver 200 ms to 500 msOver 500 msResponsiveness across qualifying interactionsPageSpeed Insights field data; Search Console
CLS0.1 or lessOver 0.1 to 0.25Over 0.25Unexpected visual movement during a visitPageSpeed Insights field data and Lighthouse; Search Console
TTFB0.8 s or lessOver 0.8 s to 1.8 sOver 1.8 sTime until the first byte of the document response arrivesPageSpeed Insights field data (Lighthouse flags a slow initial server response as a diagnostic)
FCP1.8 s or lessOver 1.8 s to 3.0 sOver 3.0 sTime until the first content is paintedPageSpeed Insights field data and Lighthouse

These thresholds are evaluated at the 75th percentile of page loads, segmented by mobile and desktop. TTFB and FCP are diagnostic metrics, not Core Web Vitals; lab measurements are useful for investigation but do not establish a field percentile. The Web Vitals definitions explain how the metrics fit together.

TermDefinitionRule it implies
Field dataMeasurements from actual visitorsCheck the device segment and whether the data describes the URL or the whole origin.
Lab dataA controlled test of a page under specified conditionsUse it to reproduce and diagnose problems; compare tests run under similar conditions.
CrUXChrome UX Report, a source of eligible Chrome visitors' field dataA public page needs sufficient popularity to appear; many small business pages may have no data.
LighthouseA lab tool that audits a page under simulated conditionsTreat its performance score as a test result, not a customer percentile.
PageSpeed InsightsA tool that presents CrUX field data when available and a Lighthouse lab auditCheck which data is for the URL and which is for the origin.
Search Console Core Web Vitals reportA field-data report that groups similar URLsInvestigate the affected group and its worst metric.
Page weightBytes transferred to load a pageReduce unnecessary bytes, especially those needed early in a visit.
Mobile-first indexingGoogle's use of mobile content for indexing and rankingKeep the mobile version's primary content available to Google's smartphone crawler.
Target sizeThe dimensions of an interactive pointer target in CSS pixelsCheck WCAG's minimum size or a permitted exception.
48%
Mobile websites passing all three Core Web Vitals
76%
Mobile pages with an image as the LCP element
2.56 MB
Median mobile home page weight
12 March 2024
Date INP replaced FID
LCP, INP, and CLS rulers showing good, needs improvement, and poor threshold bands

Each ruler uses a different unit. The field-data rating comes from the 75th percentile, so a good result means at least 75% of measured loads meet the good threshold.

LCP: Largest Contentful Paint

Definition. LCP measures when the largest eligible content element visible in the initial viewport finishes rendering. It is a useful approximation of when the main content has appeared, although it does not promise that every part of the page is ready.

Candidates include an img, an image inside an SVG, a video poster or first frame, a CSS url() background image, and block-level text. CSS gradients do not count as LCP image candidates.

Where to see it. PageSpeed Insights shows LCP in its field data when enough CrUX data exists and reports a lab LCP in its Lighthouse audit. The audit can identify the element selected during that test; inspect the actual element before deciding which image or text block to optimize.

Search Console shows LCP issues for groups of URLs with sufficient field data. A group may contain a shared template problem even when one tested URL happens to load quickly.

Threshold or rule. Good LCP is 2.5 seconds or less at the 75th percentile; over 4.0 seconds is poor. The Web Almanac found that 62% of mobile websites had good LCP and that an image was the LCP element on 76% of mobile pages, using July 2025 data.

The trap. A hero image is a sensible first suspect, but it is not automatically the measured LCP element. A large heading may win, or the selected element may differ between mobile and desktop layouts. Check the element and its loading chain before changing image settings.

INP: Interaction to Next Paint

Definition. INP measures responsiveness to qualifying interactions across a visit, reflecting how long users wait for the next visual update after an interaction. It observes clicks, taps, and key presses, including input from an onscreen keyboard; hovering, scrolling, and zooming are excluded.

INP replaced First Input Delay (FID) as a Core Web Vital on 12 March 2024. FID was removed from Search Console that day, so an old report recommending FID as the current responsiveness target needs updating.

Where to see it. Look for field INP in PageSpeed Insights or the Search Console Core Web Vitals report. A page without enough qualifying visitor interactions may lack an INP value even when you can test its buttons yourself.

For a lab investigation, use browser developer tools to record an interaction and inspect work on the main thread. Test a menu, a form field, and the action that moves a customer to the next step.

Threshold or rule. An INP of 200 milliseconds or less is good; over 500 milliseconds is poor. The Web Almanac reported good INP for 77% of mobile websites and 97% of desktop websites in July 2025, its largest mobile-to-desktop gap among the three metrics.

Break up expensive JavaScript work and remove scripts that do not earn their place. If the form itself asks for too much or has confusing steps, use the conversion path design guide to examine the journey as well as the response time.

The trap. A button can respond quickly on a powerful laptop while a phone waits for JavaScript to finish. A single successful click also tells you little about the range of interactions visitors make during a full visit.

CLS: Cumulative Layout Shift

Definition. CLS measures unexpected movement of visible elements. Its score is the largest session window of layout shifts: shifts must be less than one second apart, and the window can last at most five seconds.

Shifts within 500 milliseconds of a tap, click, or key press are excluded when they qualify as recent user input. A menu opening after a tap is therefore treated differently from an image arriving late and pushing a button down.

Where to see it. PageSpeed Insights provides field CLS when available and a lab CLS for the tested load. Use browser developer tools to inspect layout shifts and identify elements that moved, then trace what caused the space to change.

Search Console can show a CLS problem across a group of URLs. Check shared elements such as banners, images, embeds, and fonts when the affected pages use the same template.

Threshold or rule. A score of 0.1 or less is good; over 0.25 is poor. In the Web Almanac's July 2025 data, 81% of mobile websites had good CLS compared with 72% of desktop websites, so desktop deserves its own check.

Give images width and height attributes so the browser can reserve their aspect ratio before the image arrives. Reserve space for other late content where its dimensions are known, and check font swaps for changes in text size or line wrapping.

The trap. A lab run sees only the activity that occurs during that run. A late banner or an embed that appears after the test may still move content for visitors, so inspect field data and exercise the page after the initial load.

TTFB and FCP: Diagnostics for Loading

Definition. Time to First Byte (TTFB) measures the time until the first byte of the document response arrives. First Contentful Paint (FCP) measures when the browser first paints content. Neither is a Core Web Vital, but both help explain a slow LCP.

Where to see it. PageSpeed Insights can show both metrics in CrUX field data when available, and FCP in its Lighthouse audit; Lighthouse flags a slow initial server response as a diagnostic. A server timing investigation and the browser's network panel can help you follow a slow document response; the Lighthouse diagnostics can point toward resources delaying the first render.

For the search-side consequences of slow responses, see crawling, rendering, and server response. A slow response is a useful lead for both a site maintainer and whoever operates the server.

Threshold or rule. Good TTFB is 0.8 seconds or less; good FCP is 1.8 seconds or less. The Web Almanac found good TTFB on 44% of mobile websites in July 2025.

A slow TTFB directs attention to the server response and delivery path. If TTFB is acceptable but FCP is slow, inspect resources that block rendering, including CSS and scripts needed before the browser can show content.

The trap. Improving TTFB does not guarantee a good LCP: the browser may still have to discover, download, and render the largest element. Treat the two diagnostic metrics as leads, then verify whether the LCP result changes.

Field Data Versus Lab Data: CrUX and Lighthouse

Definition. Field data describes visits by real users under their actual conditions. Lab data comes from a controlled run, making it useful for reproducing a problem and comparing changes.

Chrome's UX Report includes public pages and origins that are sufficiently popular; Google does not disclose the eligibility threshold, and a page cannot be submitted for inclusion. Many small business pages may therefore have no field data.

Where to see it. CrUX supplies the field data shown in PageSpeed Insights and the Search Console Core Web Vitals report. According to the CrUX methodology, contributors are eligible Chrome users who have enabled usage statistics and meet the other participation conditions, on desktop or Android.

Chrome on iOS, Android WebView, and other Chromium browsers such as Edge do not contribute. iPhone visitors are therefore invisible to CrUX, an important limitation if they make up much of your audience.

Threshold or rule. Lighthouse's default mobile profile uses simulated throttling with 150 milliseconds of round-trip time, 1.6 Mbps download, 750 Kbps upload, and a fourfold CPU slowdown. The Lighthouse throttling documentation describes the simulation; the CPU multiplier is relative to the machine running the test.

Use a consistent lab setup when comparing changes. A real phone on mobile data adds a useful check of an actual device and connection, although one phone visit is still a sample of one.

The trap. An unthrottled developer laptop on a fast office connection can make the same page feel much quicker than it does on a customer's phone. Even simulated Lighthouse results can vary with the host machine, so record the test setup when comparing runs.

Developer laptop and customer phone compare the same page on fast and weak connections

Notice the difference in devices and connections: the laptop test can hide delays a customer encounters on a phone. A controlled lab profile helps expose those delays, while field data describes eligible real visits.

A green Lighthouse score is a lab result

A score of 90 or more describes one Lighthouse audit under its test conditions. It does not establish that your customers have good Core Web Vitals at the 75th percentile.

PageSpeed Insights and the Search Console Report

Definition. PageSpeed Insights combines a Lighthouse lab audit with CrUX field data when that data is available. Its field view uses a 28-day collection period and may describe the requested URL or fall back to its origin.

The Search Console Core Web Vitals report uses field data and groups similar URLs. It can flag a group because its worst metric is poor even when another metric in that group is good.

Where to see it. Enter a URL in PageSpeed Insights and read the field-data label carefully: it tells you whether the data describes the URL or the origin. In Search Console, open the separate mobile and desktop Core Web Vitals reports, then inspect the affected URL groups and issue types.

The Search Console report documentation explains grouping and validation. A group needs sufficient data for both LCP and CLS to appear, so absence from the report is not a performance certificate.

Threshold or rule. Lighthouse performance scores use bands of 90 to 100 for good, 50 to 89 for needs improvement, and 0 to 49 for poor. Those scores are lab results; the Core Web Vitals thresholds apply to the individual field metrics at the 75th percentile.

After a fix, Search Console validation runs a 28-day monitoring period. Allow that period to gather evidence across the group rather than expecting a single new lab run to clear the issue immediately.

Use this decision rule when you open PageSpeed Insights:

  1. URL field data is present: use it to assess that URL's real-user measurements for the selected device segment.
  2. Only origin data is present: use it as context for the site, recognizing that it aggregates visits across pages and may hide a template-specific problem.
  3. No field data is present: use the lab audit and test on a real phone over mobile data. Do not claim a 75th-percentile result from either test.

The trap. PageSpeed Insights and Search Console can appear to disagree because one view may show an individual URL or origin alongside a lab score, while the other reports grouped URLs and their worst field metric. Compare the device, data scope, metric, and collection period before treating the difference as a fault.

Page Weight: Bytes a Page Transfers

Definition. Page weight is the amount of data transferred to load a page, including resources such as images and JavaScript. Weight is not a Core Web Vital, but bytes consume connection capacity and can delay resources the browser needs early.

Where to see it. Use the browser's Network panel to inspect transferred bytes by resource type. Test a first visit with cache disabled, then compare a repeat visit with cache available; those are different experiences.

The Web Almanac page-weight chapter reports that the median home page in its July 2025 crawl was 2.86 MB on desktop and 2.56 MB on mobile. Its mobile figures include about 911 KB of images and about 632 KB of JavaScript.

Threshold or rule. There is no Core Web Vitals pass mark for total page weight. Set a budget for the resources your page needs, then check whether the budget protects the content and interactions customers use first.

For an openly hypothetical calculation, assume all 2.56 MB must cross a connection at Lighthouse's simulated 1.6 Mbps download rate:

Hypothetical minimum transfer time for all page bytes

2.56 MB * 8 / 1.6 Mbps = 12.8 seconds

That is a bandwidth floor for moving every byte under the stated assumption, not a prediction of LCP or a complete load time. Latency adds time, LCP can happen before every byte arrives, and cached resources can help repeat visits.

The trap. A small total does not guarantee a fast page if a critical resource is discovered late or the server responds slowly. When the template itself is the bottleneck, Iron Goo's Foundation site rebuild specifies a 90 plus mobile Lighthouse score and performance budgets in CI; the Lighthouse score is a lab result.

Images: Delivery, Dimensions, and Priority

Definition. Image optimization means sending an appropriate image file and size, reserving its space, and making sure the browser discovers important images promptly. The image selected as LCP deserves particular attention because delaying it can delay the loading metric.

Where to see it. In PageSpeed Insights, inspect the LCP element and image-related Lighthouse diagnostics. In the Network panel, check which image file was requested, when the request started, and whether the browser downloaded a much larger image than the displayed size requires.

Threshold or rule. Use AVIF or WebP where suitable, and provide responsive choices with srcset and sizes. Include dimensions to reserve space. Use loading="lazy" for images below the initial viewport, and do not lazy-load the LCP image.

The following hypothetical markup gives the hero a high fetch priority and leaves a later image eligible for lazy loading. Replace the paths, dimensions, and alt text with values appropriate to your own images.

<img src="/img/hero-1200.avif"
     srcset="/img/hero-600.avif 600w, /img/hero-1200.avif 1200w"
     sizes="100vw"
     width="1200"
     height="675"
     fetchpriority="high"
     alt="A technician inspecting a control panel">

<img src="/img/team.avif"
     loading="lazy"
     width="800"
     height="533"
     alt="Team members meeting in a workshop">

The Fetch Priority example reports a Google Flights LCP improvement from 2.6 seconds to 1.9 seconds. That result is an example, not a promised saving; priority also does not make an undiscovered image discoverable.

The trap. The Web Almanac found that about 16% to 17% of pages still lazy-loaded their LCP image in its 2025 data. Applying lazy loading to every image can delay the very image needed for the initial view. A CSS background image may need earlier discovery, such as a preload, as well as an appropriate priority.

Web Fonts: Loading and Fallback Movement

Definition. A web font is a font file downloaded for display in the browser. WOFF2 is a suitable compressed format; font-display: swap lets fallback text appear while the requested font loads.

Where to see it. In the Network panel, filter for fonts and check which files the page requests. Watch the heading and body text during a throttled load, then inspect any layout shift associated with the font changing.

Threshold or rule. Preload only the one or two font files actually used above the fold when early loading is needed. Check the exact weights and styles used there so the preload points to files the page will use.

Choose fallback fonts with similar metrics where practical. Differences in character width or line height can change wrapping and move neighboring content when the web font replaces the fallback.

The trap. font-display: swap makes text visible sooner, but it can still contribute to CLS. Preloading every font weight can also compete with more important early resources, including the LCP image.

Third-Party Scripts: Requests and Main-Thread Work

Definition. Third-party scripts come from services outside your site's own application, such as analytics tools, tag managers, and chat widgets. Their downloads add requests, and their JavaScript can compete for time on the browser's main thread.

Where to see it. Review the Network panel for third-party requests and use a performance recording to inspect main-thread work around slow interactions. Open the tag manager and list every published tag, its purpose, and who uses its output.

The Web Almanac's 2025 third-party chapter found that roughly 90% of pages used at least one third party. The median mobile page made 79 third-party requests.

Threshold or rule. Remove a tag when nobody reads its output and there is no other reason to keep it. For each remaining tag, check whether it needs to run on every page and whether it needs to run before the customer can interact.

Retest the affected interaction after each meaningful change. A tag manager is a delivery mechanism, so inspecting only the tag manager's own request will not reveal the full list of scripts it starts.

The trap. A small script file can trigger more network requests or run substantial work after it loads. Total transferred bytes alone will not tell you how much main-thread time an analytics tag or chat widget consumes.

Viewport and Tap Targets

Definition. The viewport setting tells a mobile browser how to size the layout viewport. A tap target is the interactive area a person can activate with a pointer, measured in CSS pixels.

Where to see it. Inspect the document head for the viewport declaration and test the page on a phone-sized viewport. Check the actual clickable area of links, buttons, and controls, including controls that sit close together.

<meta name="viewport" content="width=device-width, initial-scale=1">

Without an appropriate viewport setting, a mobile browser may lay out the page at a desktop-like width and scale it down. Text and controls can then appear much smaller than intended.

Threshold or rule. WCAG 2.2 Success Criterion 2.5.8, Target Size (Minimum), is Level AA and generally requires pointer targets of at least 24 by 24 CSS pixels. The criterion includes exceptions, including a spacing exception; check the actual conditions in WCAG 2.2.

For the rest of the standard and a wider review of barriers, use the accessibility and inclusive design guide. A control can meet the minimum size and still be difficult to identify or operate.

The trap. Measuring only the visible icon can misstate the target size: inspect the full interactive area. Conversely, extra-looking space inside a visual design does not help if it is outside the clickable element and the applicable spacing exception is not met.

Responsive Sites and Mobile-First Indexing

Definition. A responsive site adapts its layout to the viewport while serving a shared site. A separate mobile site can use different URLs, commonly an m subdomain, and requires the two versions to be maintained consistently.

Google uses the mobile version of a site's content, crawled with its smartphone agent, for indexing and ranking. That makes missing mobile content a search issue as well as a customer experience issue.

Where to see it. Compare the primary content on mobile and desktop, including text, images, and important links. Inspect the mobile version's source and loading behavior, and check the canonical and alternate relationships if you operate separate URLs.

Threshold or rule. Google recommends responsive web design as the easiest pattern to implement and maintain. For a separate mobile site, keep the same primary content on both versions and use the desktop URL as canonical, with the mobile URL identified as the alternate.

Do not require a swipe, click, or typing action to lazy-load primary content for Google's crawler. Content can be available on mobile even when its layout differs; verify availability rather than demanding identical visual arrangements.

The trap. A mobile layout that looks complete at first glance may omit copy or links that appear on desktop. Check the content itself and how it loads, particularly on a separately maintained mobile site.

Offline and Poor Networks

Definition. A poor network can delay requests, interrupt submissions, or leave a customer unsure whether an action worked. Offline behavior describes what the site does when the connection is unavailable.

Where to see it. Use the Network panel in browser developer tools with throttling presets such as Slow 4G and Offline. On a test form, type a recognizable sample, interrupt the connection, and submit; check the visible message and whether the entered values remain.

Also test on a real phone over mobile data. The combination gives you a repeatable failure condition and a practical view of the form on a device a customer might use.

Threshold or rule. A failed submission should say clearly that it failed and keep what the person typed so they can retry. Check that a retry after the connection returns has an understandable outcome.

A service worker can cache an app shell or provide an offline page. In my opinion, an offline page is optional for most small sites; preserving form input and explaining a failed submission is the more useful first requirement.

The trap. A spinner that stops without a message leaves the customer to guess whether the form arrived. A generic error that clears the fields forces them to recreate their work, so test both the message and the retained values.

What Speed Is Worth in Conversions

Definition. A conversion result measures a customer action, such as reaching a form submission page or completing a booking. Speed may affect that action, but a percentage reported for one group of sites is not a universal forecast for yours.

Where to see it. The Milliseconds make millions study was commissioned by Google and conducted by 55 and Deloitte. It observed mobile visits across 37 large European and American brand sites, with more than 30 million sessions over 30 days at the end of 2019.

Threshold or rule. The study associated a 0.1-second improvement across four speed metrics with 21.6% more users reaching the form submission page for lead generation and a 10% increase in booking rates for travel. The four metrics were First Meaningful Paint, Estimated Input Latency, Observed Load, and Max Server Latency or TTFB.

Those are observed associations for the study's conditions, not guaranteed effects of changing one Core Web Vital. Use your own form and booking measurements to judge whether a speed change coincides with a useful business result.

The trap. The study was observational, covered large brands, and measured mobile visits; it was not a randomized test proving the same effect for every company. A generic claim that every second costs a fixed percentage of conversions, without a source and comparable conditions, is not worth repeating.

Free ebook

Put your next steps in order

Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.

By signing up and confirming your address, you consent to receive the free ebook plus one email every two weeks with 3 or 4 new Iron Goo videos, guides and articles, and nothing else; unsubscribe any time with one click.

Related in UX

Ready to move?

Tell us where your business is today. A written assessment lands in your inbox within two business days.

Talk to us