Core Web Vitals fail for specific engineering reasons, not as a single score: LCP = critical rendering path, INP = main-thread blocking, CLS = missing size reservation.

My checklist

1. LCP: preload hero image with fetchpriority="high", compress to AVIF/WebP, inline critical CSS, defer non-critical JS.
2. INP: break long tasks (setTimeout / requestIdleCallback), move heavy work to Web Workers, cut bundle size.
3. CLS: set explicit width/height on images, use aspect-ratio in CSS, font-display: swap with metric-matched fallbacks.
4. Verify: lab (Lighthouse) + field (web-vitals lib reporting to analytics), monitor regressions in CI.

Where audits fit

Manual checks miss cross-issue correlation (e.g. a font causing both LCP delay and CLS). I use automated audits to rank fixes by impact before touching code. Disclosure: I'm affiliated with Scanly. A sample report structure is here: https://scanly.site/sample-report

What do you prioritize first when all three vitals fail on the same page?