Largest Contentful Paint: usually an image problem
The single most common LCP fix across our audits has been the hero image - either too large a file, not preloaded, or lazy-loaded when it shouldn't be, since the largest above-the-fold element should never be deferred:
<link rel="preload" as="image" href="hero.webp">
<img src="hero.webp" fetchpriority="high">
Cumulative Layout Shift: reserve the space
img, iframe { aspect-ratio: attr(width) / attr(height); }
Almost every CLS problem we've found traces back to an image or an ad slot with no reserved dimensions, causing content to jump once it finally loads.
Interaction to Next Paint, the newer signal
Heavy JavaScript running on the main thread during an interaction - a large synchronous computation triggered by a click - is the usual culprit; breaking that work into smaller chunks keeps the interface responsive.
Cumulative Layout Shift was the one nobody was watching
LCP and load speed had years of tooling and attention behind them by 2022; CLS was the metric that quietly wrecked scores on otherwise fast sites, usually from images without dimensions, web fonts causing a layout jump (FOUT/FOIT), or an ad slot reserving no space until its content loaded in.
<img src="hero.jpg" width="1200" height="630" alt="…">
/* explicit dimensions reserve the space before the image loads */
What actually moved the needle for us
Setting explicit width/height (or aspect-ratio) on every image and embed fixed the majority of CLS issues we found across our sites without touching a line of JavaScript. For LCP, self-hosting fonts with font-display: swap and preloading the hero image did more than any bundle-size optimization. The lesson that stuck: Core Web Vitals rewards fixing specific, identifiable layout mistakes far more than it rewards general "make everything faster" efforts.