Core Web Vitals: an evidence-based measurement and prioritisation framework for SME websites

Core Web Vitals are user-experience metrics that describe real-world loading performance, interactivity and visual stability. They sit inside the broader page-experience discussion and are used by Google’s ranking systems – but they do not replace relevant content, clean indexing or trustworthy business signals.

Three metrics form the current core: Largest Contentful Paint (LCP) for perceived load, Interaction to Next Paint (INP) for responsiveness, and Cumulative Layout Shift (CLS) for visual stability. INP replaced FID as the responsiveness Core Web Vital; optimising only FID means working with a retired core signal.

web.dev defines target bands for a good experience and recommends looking at the 75th percentile of page loads, segmented by mobile and desktop. That means the best lab score is not the goal; the majority of real visits landing in the “good” band is.

Common thresholds: LCP good up to 2.5 s, needs improvement below 4.0 s, poor above. INP good up to 200 ms, needs improvement below 500 ms. CLS good up to 0.1, needs improvement below 0.25. These are documented research thresholds – not a guarantee of any specific ranking.

Lab tools (Lighthouse, WebPageTest, local builds) are diagnostic: they make causes reproducible under controlled conditions. Field data (Chrome UX Report, Search Console) describes what real users experience – including weak networks, mid-range phones and cache states.

A common mistake is celebrating a lab score of 90 while mobile 75th-percentile LCP sits at 4 s. Conversely, a harsh lab profile can anticipate field pain before a site has enough traffic for CrUX. For SMEs with low traffic: run lab rigorously; adopt field as soon as it is available.

Chrome and web.dev guides on LCP and INP encourage segmenting causes rather than “optimising everything at once”. For LCP, typical drivers include late discovery of the LCP element, render-blocking resources, slow server response, heavy client work and late image delivery. For INP: long tasks, excessive script work and expensive event handlers.

For mid-market sites a tight focus works: (1) paint hero content without client wait states, (2) defer non-critical JavaScript, (3) limit fonts and images, (4) reserve layout for media and consent UI. Only then chase micro-optimisations such as residual unused CSS.

Scientific seriousness in practice means reproducibility and clear sources – not a lab coat. For every run, record URL, device/profile, throttling, tool version, commit or deploy ID, and whether it was a first visit or warm cache. Only compare like with like.

External claims such as “Google only ranks pages above 90” do not belong in serious advice. What can be evidenced: Core Web Vitals and page experience are relevant quality and ranking signals; helpful content and technical crawlability remain foundational. Both layers belong together.