Core Web Vitals: optimize LCP, INP and CLS before changing hosting
Published on 2026-11-14 by frfrand
How to diagnose LCP, INP, and CLS with field and lab data, fix the frontend, and detect server bottlenecks.
# Core Web Vitals: optimize LCP, INP and CLS before changing hosting
Changing servers can improve latency, but it does not automatically fix a 4 MB hero image, blocking JavaScript, or elements that jump while loading. Before buying another infrastructure, identify which part of the experience is failing.
1. What each indicator measures
| Metric | What it measures | Good reference value | | --- | --- | --- | | LCP | Appearance of main content | 2.5 s or less | | INP | Interaction latency | 200 ms or less | | CLS | Visual instability | 0.1 or less |
Objectives are normally evaluated at the 75th percentile of real visits, separating mobile and desktop. They are experience-oriented thresholds, not a ranking guarantee.
2. Field versus lab
Consult PageSpeed Insights and the Core Web Vitals report in Search Console for real data when there is sufficient volume. Use Chrome DevTools and Lighthouse to investigate causes in a reproducible environment. Do not confuse a lab-obtained Lighthouse score with real user experience. Measure representative routes: home, product page, search, and checkout.
3. Improve LCP without changing hosting
Identify the LCP element: banner, title, or image. If it is an image, use WebP/AVIF when appropriate, responsive variants, compression, explicit dimensions, and avoid lazy loading on the main visible image. Reduce the time it takes for the browser to discover it; fetchpriority="high" can help for a truly priority hero resource. Optimize fonts and critical CSS. Check TTFB, but remember that LCP combines more than server response.
4. Improve INP
An interface with many listeners, heavy JS, or long tasks takes time to respond. In DevTools Performance, look for blocking tasks, costly re-renders, and slow handlers. Split long tasks, defer third parties, reduce unnecessary hydration, and use virtualization for large lists. In React, measure before applying massive memorizations: optimizing without profiler can complicate the code without benefit.
5. Improve CLS
Reserve width and height for images, embeds, and ads. Avoid inserting banners over already visible content and choose a font strategy that reduces unexpected movements. Test real navigation, not just initial load.
6. When to review the server
Analyze TTFB by region, cache, slow queries, CPU, RAM, worker saturation, and proximity to the user. If HTML takes too long to generate, database improvements, caching, or hosting can have a real impact. If TTFB is good and the blockage is in JavaScript or images, changing VPS probably won't solve the cause.
7. Work plan
Save baseline, identify affected template, formulate a hypothesis per change, deploy to staging, measure again, and wait for new field data. Prioritize improvements that raise experience and conversions, not just scores.
Complement with SSR vs SSG and shared hosting, VPS or dedicated.
Frequently asked questions
Does Lighthouse 100 guarantee good SEO? No. Does a VPS fix INP? Usually INP depends largely on work in the browser. Is a single URL optimized? Start with a template and validate its impact on similar routes.
Source: https://web.dev/articles/vitals/