Website Speed Statistics 2026: 5,987 Websites Measured
By Christian Schreiber · Updated
PerfDelta measured the home pages of 5,987 websites, mostly well-known brands from the US, the UK, Germany, France and the Netherlands. The figures come from field data of real Chrome users (CrUX) and from one Lighthouse lab test per home page. Everything on this page is recalculated daily, and one snapshot per month is kept as history.
Key statistics
- 48% of the measured websites pass all three Core Web Vitals with real-user data. Across the full CrUX dataset the figure is 56% (May 2026).
- INP is the metric most sites fail: only 73% have a good INP, compared with 70% for LCP and 74% for CLS.
- The median mobile Lighthouse score is 43. Only 2% of home pages score 90 or higher.
- 50% of the websites that pass the Core Web Vitals still have a mobile lab score below 50.
- The median home page weighs 4.3 MB, of which 1.5 MB is JavaScript. 84% are heavier than 2 MB.
- 96% of home pages load third-party code, from a median of five providers. Google Tag Manager is on 67%.
- On 62% of home pages the largest visible element (LCP) is an image.
- The most common lab finding is render-blocking resources, affecting 91% of home pages.
Core Web Vitals pass rate by metric
Share of the 5,748 websites with field data that meet each threshold at the 75th percentile. Measured across all devices.
| Metric | Good | Median | Middle half | Threshold |
|---|---|---|---|---|
| LCP | 70% | 1977 ms | 1499 ms – 2696 ms | 2500 ms |
| INP | 73% | 144 ms | 102 ms – 209 ms | 200 ms |
| CLS | 74% | 0.04 | 0.01 – 0.11 | 0.10 |
For comparison: in the May 2026 CrUX release, 55.9 percent of roughly 18.4 million origins pass all three metrics, with 68.6 percent good LCP, 86.6 percent good INP and 81.3 percent good CLS (Chrome UX Report, May 2026). Large brand websites do markedly worse on INP than the web as a whole. More context in the Core Web Vitals benchmark.
Failed Core Web Vitals by metric
Breakdown of the websites that miss at least one threshold.
-
two or three metrics23%
-
only INP too slow9%
-
only LCP too slow11%
-
only CLS too high9%
A site that fails a single metric has a well-defined problem. 23% of websites fail more than one. For those sites, page weight and third-party code further down this page are the place to look.
Lab scores and real user experience diverge
The mobile Lighthouse score simulates a slow phone on a throttled connection. Its median is 43, and 62% of home pages score below 50. Many of these sites still pass the Core Web Vitals, because their real visitors use faster devices and networks. Rankings use the field data. The lab score is a diagnostic tool, not a verdict.
50% of websites that pass the Core Web Vitals have a lab score below 50.
More metrics: TTFB, FCP, TBT and Speed Index
TTFB and FCP from real-user field data (mobile, 75th percentile), TBT and Speed Index from the mobile lab test. The "good" threshold follows Google (TTFB, FCP) and Lighthouse (TBT, Speed Index).
| Metric | Source | Good | Median | Middle half | Threshold |
|---|---|---|---|---|---|
| Time to First Byte | Field data | 61% | 693 ms | 490 ms – 1009 ms | 800 ms |
| First Contentful Paint | Field data | 63% | 1545 ms | 1179 ms – 2112 ms | 1800 ms |
| Total Blocking Time | Lab | 24% | 680 ms | 214 ms – 1531 ms | 200 ms |
| Speed Index | Lab | 5% | 8657 ms | 6039 ms – 12347 ms | 3400 ms |
Page weight: JavaScript is the largest share
Bytes transferred for the home page in the mobile lab test, median by resource type.
-
JavaScript1.5 MB
-
Images1.1 MB
-
Fonts173 KB
-
CSS91 KB
-
Other84 KB
-
HTML65 KB
The median home page transfers 4.3 MB in 122 requests, with the middle half between 2.5 MB and 7.5 MB. 45% of home pages are heavier than 5 MB.
Third-party code on 96% of home pages
Share of home pages that load the provider, and the median main-thread time its code takes in the lab test.
| Provider | Share | Main thread (median) |
|---|---|---|
| Google Tag Manager | 67% | 430 ms |
| Meta (Facebook) | 27% | 283 ms |
| OneTrust (Optanon) | 17% | 118 ms |
| Google Ads (DoubleClick) | 16% | 153 ms |
| Google CDN | 12% | 321 ms |
| Other Google APIs and SDKs | 11% | 44 ms |
| Google Analytics | 11% | < 1 ms |
| TikTok | 9% | 168 ms |
| Microsoft Clarity | 9% | 84 ms |
| Google Fonts | 9% | < 1 ms |
| Adobe Tag Manager | 7% | 145 ms |
| JSDelivr CDN | 5% | 21 ms |
| Cloudflare | 4% | 9 ms |
| YouTube | 4% | 585 ms |
| Cloudflare CDN | 4% | 21 ms |
Every line of third-party code competes with the page itself for the main thread. That mostly hurts INP, the response to clicks and input. The guide to Interaction to Next Paint explains how to fix it.
The largest element: usually an image, usually late
On 62% of home pages the LCP element is an image or video. Splitting the lab LCP time into its parts, render delay is the largest part on 51% of pages and resource load delay, the wait before the image is even requested, on 42%. The download of the image itself is almost never the bottleneck.
| Part of LCP time | Median | Largest part on |
|---|---|---|
| Time to first byte | 16 ms | 1% |
| Resource load delay | 803 ms | 42% |
| Resource load duration | 108 ms | 6% |
| Element render delay | 873 ms | 51% |
Most common lab findings
Share of home pages with the finding and the median estimated saving according to Lighthouse.
| Finding | Share | Saving (median) |
|---|---|---|
| Render-blocking resources | 91% | 5232 ms |
| Unused JavaScript | 86% | 1650 ms |
| Images not properly sized | 74% | 1663 ms |
| Unused CSS | 68% | 450 ms |
| Images not in next-gen formats | 49% | 2152 ms |
| Unminified JavaScript | 14% | 300 ms |
| Unminified CSS | 9% | 150 ms |
| Text compression missing | 5% | 437 ms |
Core Web Vitals by industry
Industries with at least 30 measured domains. Each has its own page with distribution, platforms and trend.
| Industry | Domains | Pass rate | LCP (median) |
|---|---|---|---|
| Public sector and non-profit | 361 | 67% | 1593 ms |
| Education | 363 | 62% | 1706 ms |
| Professional services | 167 | 60% | 1747 ms |
| Media and publishing | 570 | 60% | 1672 ms |
| Real estate | 192 | 58% | 1862 ms |
| Trades and local services | 98 | 56% | 1840 ms |
| Software and SaaS | 317 | 55% | 2093 ms |
| Health and medical | 282 | 55% | 1921 ms |
| Logistics and transport | 151 | 48% | 1817 ms |
| Manufacturing and industry | 175 | 45% | 2235 ms |
| Finance and insurance | 527 | 44% | 2168 ms |
| Entertainment and leisure | 339 | 41% | 2058 ms |
| Food and restaurants | 273 | 41% | 2233 ms |
| Energy and utilities | 165 | 41% | 2185 ms |
| Online retail | 1110 | 40% | 2001 ms |
| Automotive and mobility | 190 | 36% | 2048 ms |
| Telecommunications | 194 | 26% | 2515 ms |
| Travel and hospitality | 247 | 23% | 2541 ms |
Core Web Vitals by CMS and platform
Platforms with at least 30 domains with field data, detected from the home page technology stack.
| Platform | Domains | Pass rate | LCP (median) | Lab score (median) |
|---|---|---|---|---|
| TYPO3 | 60 | 85% | 1510 ms | 61 |
| Webflow | 40 | 73% | 2135 ms | 39 |
| Drupal | 356 | 61% | 1818 ms | 46 |
| Gatsby | 30 | 60% | 1994 ms | 47 |
| Storyblok | 43 | 58% | 1767 ms | 50 |
| Shopify | 293 | 57% | 1836 ms | 41 |
| Sanity | 54 | 54% | 1898 ms | 38 |
| Sitecore | 128 | 52% | 2096 ms | 44 |
| WordPress | 636 | 51% | 2053 ms | 49 |
| Fourthwall | 68 | 50% | 1945 ms | 36 |
| Next.js | 696 | 40% | 2062 ms | 38 |
| Contentful | 184 | 36% | 2118 ms | 40 |
| Adobe Experience Manager | 434 | 34% | 2336 ms | 39 |
| Nuxt | 88 | 33% | 2177 ms | 41 |
| Salesforce Commerce Cloud | 91 | 29% | 2154 ms | 33 |
| Magento | 30 | 27% | 2440 ms | 36 |
The platform sets the frame but does not decide alone. This sample consists almost entirely of large websites with their own development teams, so the gaps between systems are smaller than across the web as a whole.
Takeaways: priorities for better Core Web Vitals
The numbers on this page point to these starting points:
- Cut back third-party code: 96% of home pages load code from other providers, and only 73% have a good INP. Every script that is removed or loads later frees up the main thread. More in the guide to Interaction to Next Paint.
- Remove render-blocking resources: the finding affects 91% of home pages, with a median estimated saving of 5232 ms. The guide to render-blocking resources shows how.
- Request the LCP image early: on 62% of home pages the largest element is an image or video, yet the download itself is the largest part of LCP time on only 6%. A preload or fetchpriority="high" acts before that. Details in the guide to Largest Contentful Paint.
The free test below shows how your own website does on these points.
Statistics by country
Countries with at least 30 measured websites. From 100 websites a country gets its own page with a comparison, third parties and findings. The samples are curated, mostly well-known brands and large websites, not a cross-section of all websites in a country.
Methodology and limits of the sample
For each domain the home page was measured, and the most recent test is used. Field data: Chrome UX Report at origin level, 75th percentile, 28 days, all devices. An origin passes when LCP is at most 2.5 seconds, INP at most 200 milliseconds and CLS at most 0.1. Lab data: Lighthouse via the PageSpeed Insights API, mobile profile. The sample consists of curated lists of well-known websites plus tests run by users. It therefore represents larger, professionally run websites rather than the web as a whole. Domain names are not published. Full details are in the methodology.
Citing these statistics
You are welcome to use these figures with attribution and a link. Suggested citation:
Source: PerfDelta, Website Speed Statistics (as of October 17, 2026), https://perfdelta.com/website-speed-statistics
FAQ
How many websites pass the Core Web Vitals?
In this sample, 48% of websites with field data pass all three metrics. Across the full Chrome UX Report, 55.9 percent of origins passed in May 2026.
How heavy is an average home page?
The median is 4.3 MB in the mobile lab test. JavaScript is the largest part at 1.5 MB.
Why is my Lighthouse score low when my Core Web Vitals pass?
Lighthouse simulates a slow phone, while the Core Web Vitals measure real visitors. In this sample, that applies to 50% of the websites that pass the Core Web Vitals.
How often are these statistics updated?
Daily. One snapshot per month is kept, so changes over time stay traceable.
Compare your website with these numbers
The free test measures your pages with the same methods and puts them next to these numbers.
Test your websiteA free account saves your tests and shows how your values change from one test to the next. Daily monitoring with field data and alerts starts with the Solo plan.