INP replaced FID in March 2024, and a good deal of the advice online still has not caught up. Here is a current, practical read on which metrics affect rankings, which affect users, and where those two lists come apart.
The three that count
LCP (Largest Contentful Paint) measures how long the main content takes to appear. Under 2.5 seconds is good. It is usually dominated by server response time, render-blocking resources and unoptimised images — in that order.
INP (Interaction to Next Paint) measures responsiveness across the whole visit, not just the first interaction the way FID did. Under 200ms is good. Long JavaScript tasks on the main thread are the usual cause.
CLS (Cumulative Layout Shift) measures unexpected movement. Under 0.1 is good. Images without dimensions, injected ads and late-loading web fonts account for most of what we see.
Where rankings and experience diverge
Core Web Vitals are a ranking factor, but a comparatively light one. A page with excellent vitals and mediocre content will not outrank a genuinely better page. Where vitals earn their keep is at the margin — between two comparable results — and in conversion rate, which is frequently the larger prize.
Field data over lab data
Lighthouse gives you lab data from a simulated device. The Chrome UX Report gives you field data from real users, and that is what actually feeds into ranking systems. Optimise against field data. A perfect Lighthouse score with failing CrUX numbers means your real users are having a worse time than your test environment suggests.
A sensible order of work
Server response time, then render-blocking resources, then image delivery, then main-thread JavaScript. Most sites get the majority of their improvement from the first two.