Website performance remains uneven across the web. In 2024, good Core Web Vitals were recorded for 43% of mobile websites and 54% of desktop websites, while the median mobile homepage loaded 900 KB of images and 558 KB of JavaScript. These figures come from separate chapters of The Web Almanac 2024, which measures real websites and reports results by device, page type, and percentile.
Table of contents
- Core Web Vitals by device and site group
- Loading speed: TTFB, FCP, and LCP
- Why images dominate LCP
- Interactivity and long tasks
- Visual stability and back/forward cache
- Page weight, requests, and asset budgets
Core Web Vitals by device and site group
Core Web Vitals combine loading, responsiveness, and visual stability. The 2024 results show a persistent desktop advantage, but the size of that advantage depends on which websites or pages are included.
Across websites in 2024, 43% of mobile sites had good Core Web Vitals when INP was used, compared with 54% of desktop sites. Looking at the top 1,000 websites, 40% were good on mobile and 54% on desktop. Among the top 10,000, the figures were 33% and 46%; among the top 100,000, they were 31% and 43%. The top 1,000,000 had 36% good mobile results and 48% good desktop results. For sites ranked 10,000,000 and beyond, 44% were good on mobile and 43% on desktop. These rankings and device comparisons are reported in The Web Almanac 2024 Performance chapter.
Page type also changes the picture. In 2024, 45% of desktop home pages and 38% of mobile home pages had good Core Web Vitals. Secondary pages performed better: 61% of desktop secondary pages and 51% of mobile secondary pages were good.
| 2024 page group | Mobile good Core Web Vitals | Desktop good Core Web Vitals |
|---|---|---|
| Top 1,000 websites | 40% | 54% |
| Top 10,000 websites | 33% | 46% |
| Top 100,000 websites | 31% | 43% |
| Top 1,000,000 websites | 36% | 48% |
| Home pages | 38% | 45% |
| Secondary pages | 51% | 61% |
The table is a set of reported percentages, not a forecast or a combined average. Each comparison retains its stated site group and page type.
Loading speed: TTFB, FCP, and LCP
The loading metrics identify different stages of the experience. Time to First Byte (TTFB) reflects the wait for the first response byte, First Contentful Paint (FCP) records when initial content appears, and Largest Contentful Paint (LCP) measures the main visible content’s loading experience.
In 2024, 42% of mobile websites had good TTFB, 40% needed improvement, and 19% were poor. FCP was good for 68% of desktop websites and 51% of mobile websites in the June 2024 measurement. LCP was good for 72% of desktop websites, with 20% needing improvement and 8% classified as poor. On mobile, 59% had good LCP, 27% needed improvement, and 14% were poor. The figures are from The Web Almanac 2024 Performance chapter.
Home pages again lagged behind secondary pages. Good LCP was recorded on 63% of desktop home pages and 53% of mobile home pages. For secondary pages, the corresponding results were 82% on desktop and 72% on mobile.
The LCP breakdown helps show where time is spent. For the 2024 good-LCP group, median p75 sub-part times in the July 2024 data were 600 milliseconds for TTFB, 350 milliseconds for image-load delay, 160 milliseconds for image-load duration, and 230 milliseconds for render delay. In the needs-improvement group, the same figures were 1,360, 720, 270, and 310 milliseconds. In the poor-LCP group, they were 2,270, 1,290, 350, and 360 milliseconds. These are median p75 sub-part measurements, not a single page’s complete loading time.
Why images dominate LCP
Images are central to the largest visible element on many pages. In 2024, an image was the LCP element on 83.3% of desktop pages and 73.3% of mobile pages. Inline images accounted for only 0.3% of desktop pages and 0.4% of mobile pages. A separate media analysis found that an image was the LCP-responsible element on 68% of mobile pages. The two measurements use their respective chapter methodologies and should not be treated as one pooled estimate. Sources: Performance | The Web Almanac 2024 and Media | The Web Almanac 2024.
Discovery and loading choices matter. In 2024, 35% of mobile websites did not have an LCP element statically discoverable in the initial document. CSS initiated the undiscoverable LCP resource on 9% of mobile pages and 11% of desktop pages, while HTML initiated it on 33% of mobile pages and 38% of desktop pages. Also, 16% of mobile websites used native or custom lazy-loading on an image-based LCP element: 9.5% used native loading=lazy, and 6.7% used a custom technique.
Priority hints became more common but remained far from universal. In 2024, 15% of mobile pages used fetchpriority=high on their LCP image, up from 0.03% in 2022. On desktop pages, JPG was the LCP image format in 61% of cases, PNG in 26%, and WebP in 7%. AVIF accounted for 0.3% of pages in 2024, compared with 0.1% in 2022. These format and implementation results are reported by The Web Almanac 2024 Performance chapter.
Interactivity and long tasks
Interaction to Next Paint (INP) shows how quickly a page responds after an interaction. In 2024, 97% of desktop websites had good INP, 2% needed improvement, and less than 1% were poor. Mobile results were lower: 74% were good, 24% needed improvement, and 2% were poor.
Good mobile INP rose from 55% in 2022 to 64% in 2023 and 74% in 2024. Desktop good INP was 95% in 2022, 97% in 2023, and 97% in 2024. The trend describes the reported shares for those years; it does not establish why the shares changed.
The 2024 RUMvision data also shows how the slower tail behaves. At the 50th percentile, presentation delay was about 20 milliseconds, while input delay and processing time remained modest. At the 75th percentile, input delay was about 50 milliseconds and processing time and presentation delay increased. At the 90th percentile, input delay reached about 150 milliseconds, while processing time and presentation delay each exceeded 100 milliseconds.
Long tasks provide another view of main-thread pressure. Median long-task duration was 90 milliseconds on desktop and 108 milliseconds on mobile in 2024. At the 75th percentile, duration was 161 milliseconds on desktop and 187 milliseconds on mobile. At the 90th percentile, it reached 331 milliseconds on desktop and 377 milliseconds on mobile. Source: Performance | The Web Almanac 2024.
Visual stability and back/forward cache
Cumulative Layout Shift (CLS) was good for 72% of desktop websites and 79% of mobile websites in 2024. Desktop results included 18% needing improvement and 11% poor; mobile results included 12% needing improvement and 9% poor. The category percentages are reported as published and may reflect rounding.
Mobile good CLS rose from 60% in 2020 to 62% in 2021, 74% in 2022, 76% in 2023, and 79% in 2024. Desktop good CLS rose from 54% in 2020 to 62% in 2021, 65% in 2022, 68% in 2023, and 72% in 2024. These historical shares come from The Web Almanac 2024 Performance chapter; the earlier values are presented as legacy reported figures, not independently verified here.
Back/forward cache can make navigation feel faster by restoring a page instead of loading it again. In 2024, 35% of the top 1,000 desktop pages were ineligible for back/forward cache because of unload handlers. Across all ranks, about 16% of desktop pages were ineligible for the same reason.
Page weight, requests, and asset budgets
The median 2024 mobile homepage loaded 18 KB of HTML, 73 KB of CSS, 111 KB of fonts, 558 KB of JavaScript, and 900 KB of images. The median mobile inner page loaded 17 KB of HTML, 77 KB of CSS, 108 KB of fonts, 582 KB of JavaScript, and 348 KB of images. The homepage requested 20 images versus 14 for the median inner page. Both desktop and mobile pages requested a median of 8 CSS assets. Sources: Page Weight | The Web Almanac 2024 and JavaScript | The Web Almanac 2024.
JavaScript requests were also substantial. The median desktop page requested 23 JavaScript files and the median mobile page 22. At the 90th percentile, those counts rose to 70 desktop files and 68 mobile files. Median JavaScript payloads were 613 KB on desktop and 558 KB on mobile, while median JavaScript responses were 620 KB on desktop URLs and 570 KB on mobile URLs.
Compression and minification were not universal. In 2024, correctly compressed text resources appeared on 70% of desktop homepages, 71% of desktop inner pages, 71% of mobile homepages, and 72% of mobile inner pages. Correctly minified CSS appeared on 62% of desktop homepages and 63% of mobile homepages. Correctly minified JavaScript appeared on 58% of desktop homepages and 59% of mobile homepages. Source: Page Weight | The Web Almanac 2024.
Image distributions show why medians alone can hide heavy pages. In 2024, the median mobile image resource was 12 KB, compared with 56 KB at the 75th percentile and 177 KB at the 90th percentile. The largest image on a median mobile page was 135 KB; it was 404 KB at the 75th percentile and 1,002 KB at the 90th percentile. More than 10 million scanned pages were included in the finding that 99.9% requested at least one image. The median mobile page contained 13 img elements, while the 90th percentile contained 56. Source: Media | The Web Almanac 2024.
These measurements point to several separate performance constraints: server response time, main-content discovery, image delivery, JavaScript execution, layout stability, and total asset volume. A page can perform well on one dimension while remaining weak on another, so the device, page type, percentile, and measurement period should stay attached to every comparison.