Regarding "4 seconds wasted" per visit: HTTP Archive also publishes real-user performance data from Google, and only 10% of desktop websites take 4 seconds or m

Regarding "4 seconds wasted" per visit: HTTP Archive also publishes real-user performance data from Google, and only 10% of desktop websites take 4 seconds or more to load. (And I think that's not the average experience but the 75th percentile.) https://httparchive.org/reports/chrome-ux-report#cruxSlowLcp https://httparchive.org/reports/chrome-ux-report#cruxSlowLcp The Google data uses Largest Contentful Paint instead of Speed Index, but the two metrics ultimately try to measure the same thing. Both have pros and cons. Speed Index goes up if there are ongoing animations (e.g sliders). LCP only looks at the single largest content element. When looking at the real-user LCP data over time, keep in mind that changes are often due to changes in the LCP definition (e.g opacity 0 elements used to count but don't any more). https://chromium.googlesource.com/chromium/src/+/master/docs/speed/metrics_changelog/lcp.md https://chromium.googlesource.com/chromium/src/+/master/docs...