Core Web Vitals in Practice

Track 3 · Lesson 5 of 6 · about 15 minutes

Core Web Vitals in practice

Speed work fails in a predictable way: someone runs Lighthouse on a fast laptop, chases the score, and ships changes that move nothing users feel. This lesson is the version that starts from what is actually measured, on whom.

Field data first, always

Google’s assessment uses CrUX — field data from real Chrome users over a rolling 28 days. Lab runs (Lighthouse, PageSpeed’s top panel) are diagnostics: reproducible, controllable, and flattering, because your machine and network are better than your median visitor’s. The workflow is field data to find the problem, lab tools to locate its cause, field data again — patiently, it lags by design — to confirm the fix. Sites without enough traffic for CrUX get no field data at all, and for them the honest posture is “probably fine, verify with lab sanity checks”.

LCP — when the main thing appears

Largest Contentful Paint is nearly always one of two stories. A slow server: the browser cannot paint what it has not received, and no image optimisation rescues a 2-second time-to-first-byte — that is caching and hosting. Or a heavy hero: the LCP element is usually a large image, so it needs to be properly sized, modern-format, discovered early (preloaded or in the initial HTML, never lazy-loaded — lazy-loading the LCP image is the classic self-inflicted wound) and served with high priority.

CLS — things jumping while you read

Cumulative Layout Shift is unreserved space: images without width and height attributes, ads and embeds that inject late, and webfonts swapping in at a different size than the fallback. The fixes are mechanical — dimensions on every image, reserved slots for dynamic content, font-display strategies that avoid late reflow. CLS is the vital most often fixable in an afternoon.

INP — does the page respond

Interaction to Next Paint replaced FID in March 2024 and measures responsiveness across the whole visit: tap to visual response, worst-ish case. High INP is almost always JavaScript — long tasks blocking the main thread, heavy frameworks hydrating, third-party scripts doing nobody knows what. The fixes are less mechanical than CLS: break up long tasks, defer what is not needed, and audit the third-party scripts — the tag manager accumulates bodies for years.

Proportion

Vitals are one signal among many, not a ranking lottery. Going from terrible to acceptable is worth real effort — it shows in bounce and conversion before it shows anywhere else. Going from good to perfect is a hobby. Spend accordingly.

Going deeper: Core Web Vitals guides · WordPress speed optimization.