Most performance audits fail because they start in the lab. Real-world data tells you which pages are slow, which devices are affected, and whether a fix actually moved the needle.
Step 1: Map the user-critical flows
Identify the 2-3 journeys that drive revenue: homepage -> service page -> contact, blog -> case study, or product -> checkout. Everything else is secondary.
- Define a "money page" list with owners and KPIs.
- Pull Core Web Vitals for those pages only.
- Document baseline LCP, INP, and CLS before changes.
Step 2: Kill render blockers
In Nuxt 4, most delays come from over-eager JavaScript and CSS. Audit your critical CSS path, split oversized bundles, and delay non-essential UI libraries.
ts
export default defineNuxtConfig({
vite: {
build: { cssCodeSplit: true }
}
})Step 3: Validate with field data
Once you ship changes, compare your real user metrics. If you only improve Lighthouse scores, you likely optimized the wrong thing.
"Performance wins are earned in the field, not the lab."