Introduced in 2020, Google's Core Web Vitals gave the entire digital industry a shared benchmark for user experience: the Largest Contentful Paint (LCP). Its 2.5-second 'good' threshold quickly became the industry's guiding North Star.
Aiming for 2.5s LCP, however, simply isn't good enough these days for optimal user experience and business success. Google based this threshold on research, but had to ensure their benchmark was achievable by most websites. Both business analysis and cognitive psychology suggest much lower thresholds for optimal results.
We at Speed Kit are now extending Google's LCP thresholds in all our dashboards by adding two additional buckets to show more details for page loads faster than 2.5s.
Modernizing LCP Thresholds
This is the new chart we are using across the board which lets you focus on improving performance beyond the 2.5s threshold:
- Instant (< 300ms): A truly seamless experience. The user feels the site is an extension of their thoughts.
- Fast (< 1000ms): The user remains in a state of flow. The interface is responsive and keeps them engaged. This is the new primary target.
- OK (< 2500ms): The Google baseline. The site is usable, but the delay is noticeable. This is table stakes, not a competitive advantage.
- Needs Improvement (< 4000ms): Frustration is likely. The user is actively waiting and considering abandoning their task.
- Poor (> 4000ms): High abandonment risk. A significant barrier to business success.
Why these thresholds
The web has evolved dramatically in the last five years. Users now have faster devices and better network connections, and their expectations for speed have risen accordingly. Simultaneously, the underlying technology of the web has advanced. Modern browsers, new protocols like HTTP/3, and sophisticated optimization techniques like predictive preloading and speculation rules have made sub-second load times not just a possibility, but an achievable goal.
We defined these new thresholds with a clear goal: to set relevant, achievable targets that maximize both UX and business success, all while extending the familiar visualizations established by Google.
The idea was to split up the “good” bucket into three buckets defining two new goals to aim for. In defining these goals we considered three main areas:
- Cognitive psychology and UX research, best summarised by Jakob Nielsen, tells us that
- 1.0 second is the limit for a user's flow of thought to remain uninterrupted, even though they notice the delay.
- 0.3 seconds relates to the speed of the human sensorimotor feedback loop (blink of an eye).
- 0.1 seconds is the limit for a user to feel that the system's reaction is instantaneous.
- E-commerce business A/B testing, which shows an increasing effect of optimisation for sites that already load below 2.5s, i.e. 100ms saved on a “good” site has a much higher impact on conversion rates.
- Real User Monitoring (RUM) data on achievable performance tells us the even pages fully rendered in the background show LCP values between 100ms and 250ms.
We settled for thresholds of 1s for fast LCP and 300ms for instant experience. While 100ms would be an ideal threshold it is not widely achievable for complex sites.
Applying the new LCP buckets we’ve outlined here has already helped us demonstrate to customers the value added by improving their already “good” LCP. We’re certain the performance community in general would benefit from it as well.
