Core Web Vitals work goes wrong in a predictable way: someone runs PageSpeed Insights, sees a red number, installs a caching plugin, and declares victory while the four-megabyte hero image continues loading on every page.
Use field data, not lab scores
The score in PageSpeed Insights is a simulation. The data Google actually uses is field data from real Chrome users, shown in the Core Web Vitals report in Search Console. Work from that, on the templates that carry your traffic, and ignore the vanity score entirely.
LCP: the biggest element, loading slowly
Largest Contentful Paint measures how long until the main content appears. On most sites the culprit is one of three things: an oversized hero image, a slow server response, or render-blocking resources delaying everything. In that order, compress and correctly size the hero image and serve it in a modern format, check your hosting response time, and defer scripts that are not needed for first paint. This is usually where the largest single gain lives.
INP: the page ignoring taps
Interaction to Next Paint measures responsiveness. Poor INP is nearly always heavy JavaScript occupying the main thread: bloated third-party tags, chat widgets, tracking scripts, page-builder overhead. Audit what actually loads, remove what nobody uses, and defer the rest. Every tag someone added for a campaign three years ago is still costing you.
CLS: things moving as you read
Cumulative Layout Shift is content jumping during load, usually from images without dimensions, ads or embeds without reserved space, or fonts swapping late. Set explicit width and height on images, reserve space for anything injected, and use a sensible font-display strategy.
The order of operations
Images first, since they are the most common LCP cause and the easiest win. Then scripts, which affect INP and often LCP together. Then layout stability, which is quick once you know where it comes from. Then hosting, if the response time is genuinely the ceiling, which for cheap shared hosting it frequently is.
What to expect
Field data updates over a rolling twenty-eight day window, so improvements take weeks to appear in Search Console even when the fixes worked immediately. Measure with lab tools for confirmation, then wait for the field data to confirm it properly. If you would rather have this diagnosed than guessed at, that is what Core Web Vitals work covers.