YouTube iframes often drive Cumulative Layout Shift, but cross-origin content limits what you can attribute and fix. Reserve the slot, use a facade or lite embed, and verify with lab plus field monitoring.
Lighthouse now reports Baseline status for web platform features on a page, including third-party scripts. How to read Widely, Newly, and Limited availability, and what to do next.
When a WordPress site is slow, buyers often point at hosting first. Across hundreds of enterprise projects, the bottleneck is usually code: plugins, queries, autoload bloat, and third-party scripts, not the server tier.
How to find which tags and widgets cost you LCP, INP, and main-thread time, then fix or fence them without stripping analytics you still need.
Resource hints tell the browser what to fetch early. Use a decision table for dns-prefetch, preconnect, preload, and prefetch; preload the LCP image when it is not discoverable in HTML; keep font preloads tight; and verify with dated lab runs.
CrUX field percentiles are not a slower Lighthouse run. They aggregate real experiences that may abandon early, start in the background, restore from bfcache, or stay on one SPA document. Here is what partial page loads include and exclude, and how to report lab versus field without mixing the clocks.
When CrUX has no data for a client site, the problem is usually eligibility, not a broken PageSpeed Insights run. Here is how Chrome UX Report eligibility works, and what to monitor until field data exists.
Lazy loading images saves bandwidth below the fold, but `loading="lazy"` on the LCP image delays Largest Contentful Paint. Here is when to use the attribute, when to omit it, and how to check lab and field results.
WordPress 6.8 ships speculative loading (prefetch by default). Plugins that prerender more eagerly can improve next-page feel, but they can also burn bandwidth and muddy first-visit Core Web Vitals. Here is how agencies choose settings without guessing.
Time to First Byte includes more than server work. Name Unattributed Navigation Overhead (UNO) in client reports so the residual after redirect, DNS, connection, and request timing is not silently blamed on origin PHP.
