Field data is the only score Google uses for ranking
Core Web Vitals as a ranking signal is calculated from CrUX field data — real Chrome users on real devices and connections — not Lighthouse. A site with Lighthouse 100 in lab conditions can fail Core Web Vitals if real users experience slower performance.
Check field data in Search Console under 'Core Web Vitals' or via the public CrUX dashboard. The 75th-percentile metric is what Google uses; if 75% of your real users hit the 'Good' thresholds, you pass.
Why we recommend exceeding the official thresholds
Google's 'Good' thresholds (LCP under 2.5s, INP under 200ms, CLS under 0.1) are minimums. In competitive categories the marginal impact of moving from 2.4s to 1.8s LCP is meaningful both for organic ranking and for conversion rate. Across measured engagements, sites that beat the threshold by 30%+ saw conversion lifts of 5–15% versus baseline.
Mobile is the binding constraint. Mobile LCP on mid-range Android devices is typically 2–3× slower than desktop. Optimise for the median Android, not the latest iPhone, if you want field data to pass at the 75th percentile.
The high-leverage performance fixes in 2026
Image optimisation remains the single largest LCP lever. Serve modern formats (AVIF, WebP), set explicit width and height attributes, lazy-load below-the-fold images, and use responsive srcset. A typical site we audit has 30–60% of its bytes in oversized images.
JavaScript bundle size is the second-largest lever, especially for INP. Avoid loading the entire React/Vue/framework bundle on landing pages that don't need it. Use code-splitting, defer non-critical scripts, and audit third-party tags — chat widgets, marketing pixels, and tag managers routinely contribute 50–200 KB each.
Self-host fonts where licensing allows, use font-display: swap with size-adjusted fallbacks to minimise CLS, and preload only the fonts above the fold. Loading 8 font weights when you use 2 is a routine finding.