Core Web Vitals 2026: The INP Deep Dive (LCP, CLS, TTFB Too)

The 2026 thresholds (memorize these)

Google ranks pages using the 75th-percentile of real-user (CrUX) data. The "good" thresholds for 2026:

• LCP (Largest Contentful Paint): ≤ 2.5 seconds
• INP (Interaction to Next Paint): ≤ 200 milliseconds
• CLS (Cumulative Layout Shift): ≤ 0.1
• TTFB (Time to First Byte): ≤ 0.8 seconds
• FCP (First Contentful Paint): ≤ 1.8 seconds (good signal, not a ranking factor)

Why INP replaced FID

FID (First Input Delay) only measured the first interaction. INP measures the worst interaction across the page session. Real users notice the worst case, not the first one.

INP captures slow handlers across hover, click, keypress — the entire interaction surface. A page that scrolls smoothly but stalls when you click a dropdown will fail INP.

LCP fixes in priority order

Most LCP problems are caused by 3 things: render-blocking resources, large hero images, and slow server responses.

• Identify your LCP element via Chrome DevTools Performance panel.
• If it is an image: serve in WebP/AVIF, add fetchpriority="high", preload via <link rel="preload">.
• If it is text: inline critical CSS, eliminate render-blocking <link rel="stylesheet"> above the fold.
• If TTFB is the bottleneck: add CDN edge caching, switch to a faster host.

INP fixes in priority order

INP regressions are almost always JavaScript-induced. The fix pattern: find the long task, break it up, or defer it off the critical interaction path.

Common causes: massive bundle of React on the homepage (split-load it), third-party analytics tags firing synchronously (move to Partytown), framework hydration that blocks input.

CLS fixes in priority order

CLS = unexpected layout shifts. Fix: reserve space for images, fonts, ads, and dynamic content before they load.

Specifically: set width/height on images, use font-display: swap with size-adjust, give ad slots fixed dimensions, avoid inserting content above existing content.

When to stop optimizing

Once your 75th-percentile CrUX numbers cross the "good" threshold for 28 consecutive days, stop. Further optimization gains diminishing return for ranking.

Reinvest the engineering time into content velocity or schema implementation. Both produce more ranking lift at that stage.

How Rivalist measures CWV

Rivalist runs PageSpeed Insights against every audited page and surfaces both field (CrUX) and lab (Lighthouse) data. The audit ships specific fixes ranked by impact and effort.