LCP, INP and CLS measure different failures with different causes. Knowing which metric is failing tells you almost exactly where to look.
Largest Contentful Paint measures how long the main content takes to appear. It fails when the hero image is unoptimised, when a render-blocking stylesheet delays paint, or when the server takes too long to respond in the first place.
Interaction to Next Paint measures responsiveness — how long the page takes to react after a user acts. It fails when long JavaScript tasks block the main thread, which is usually third-party scripts or an over-eager hydration step.
Cumulative Layout Shift measures visual stability. It fails when images and embeds have no declared dimensions, when fonts swap and reflow text, or when content is injected above what the user is already reading.
Time to First Byte sets the ceiling for everything else. If the server takes 800ms to respond, no amount of front-end tuning will produce a good Largest Contentful Paint.
Check caching first: are anonymous responses cached, is there a CDN in front, and does a cold cache produce an acceptable response? Then look at the application — uncached database queries on every request are the usual cause.
font-display: swap with a well-matched fallback to limit reflow.These four changes account for the majority of improvement on a typical business site, and none of them requires re-architecting anything.
Laboratory tools are useful for diagnosis and misleading as a scorecard, because they run on a fast connection from a nearby location. What counts is field data from actual visitors on actual devices.
Use the Chrome User Experience Report or your own real-user monitoring, look at the 75th percentile rather than the average, and segment by device. A site that is comfortable on desktop and failing on mid-range Android phones is the normal finding.
When several metrics fail at once, the order of work matters because the fixes interact. Start at the server: reduce time to first byte through caching and query work, because every other metric inherits that delay.
Then address the largest contentful element directly — usually a hero image or a web font. Serve it at display size in a modern format, preload it, and remove anything render-blocking ahead of it. This single step commonly resolves LCP outright.
Only then move to interaction and layout stability, which are mostly about deferring or removing JavaScript and declaring dimensions. Working in this order avoids the common trap of optimising images on a site whose real problem is a four-second server response.
They are part of Google's page experience signals, which act as a tiebreaker rather than a primary ranking factor. Content relevance still dominates. The stronger argument for fixing them is commercial: slow, unstable pages measurably reduce conversion.
Google's thresholds are 2.5 seconds for LCP, 200 milliseconds for INP and 0.1 for CLS, measured at the 75th percentile of real users. Meeting all three on mobile is a reasonable target for a business site.
Because laboratory tools simulate one device on one connection, while field data aggregates real visitors over time. Use laboratory tools to diagnose and field data to judge.
Keep reading
Most headless CMS comparisons focus on features every product has. The decisions that matter are about content modelling, editorial experience and exit cost.
The single most common commercial mistake in a replatform is losing the URLs that already bring in traffic. A redirect map is cheap insurance.
A PWA avoids app store distribution and gives one codebase. It also gives up some platform capability. Here is how the trade actually falls.