Core Web Vitals are user-experience measurements. They are useful because they make problems visible, but a perfect laboratory score is not the goal. The goal is a site customers can read, use and trust.
LCP: when the main content appears
Largest Contentful Paint measures when the largest visible image or text block finishes rendering. Google's recommended good threshold is 2.5 seconds or less at the 75th percentile of visits. A giant hero image, slow server response or blocking stylesheet can delay it.
Start by compressing and correctly sizing the above-the-fold image, removing unnecessary font weights and checking server response times. The first hero image should not be lazy-loaded if it is the primary LCP candidate.
INP: how quickly the page reacts
Interaction to Next Paint measures responsiveness across a user's interactions. A good target is 200 milliseconds or less at the 75th percentile. Heavy JavaScript work can make buttons feel broken even when the visuals have loaded.
Break up expensive tasks, reduce redundant event handlers and avoid initializing every widget at startup. Test the actual actions your customers use: opening a menu, searching the blog and submitting a contact form.
CLS: how much the layout jumps
Cumulative Layout Shift tracks unexpected movement of visible content. A score of 0.1 or less is the recommended good threshold. Images without dimensions, late-inserted banners and font swaps are common causes.
Set width and height or aspect ratios on images, reserve space for interface elements and avoid adding unexpected content above the reader's current position.
Use field data and lab tests together
PageSpeed Insights may show real-user field data when enough traffic exists, and laboratory tests for a simulated run. Neither single run predicts every customer's experience. Segment by mobile and desktop, and inspect representative page templates rather than only the homepage.
| Metric | Recommended good result | First place to look |
|---|---|---|
| LCP | 2.5 seconds or less | Hero image, server response |
| INP | 200 ms or less | Long JavaScript tasks |
| CLS | 0.1 or less | Unreserved image/ad space |
Prioritize what moves customers
Fix broken functionality and page access before chasing a score from 92 to 98. Then focus on the slowest important templates: product landing page, quote form, pricing and client login. Measure before and after and keep the changes under version control.
Read the official Web Vitals definitions for current thresholds and measurement practices, and use the monthly maintenance checklist to keep performance from drifting.
How to run a meaningful performance test
Choose three important page types: the homepage, the page where a visitor enquires about a service and one article or app landing page. Test a mobile profile, not only a fast desktop connection. Record the URL, date, device conditions and the largest visible issue. Repeat the test after each change so you can tell whether the fix helped.
In the browser, note what actually feels wrong: does the main banner arrive late, does the menu react slowly or does text shift under a finger? LCP, INP and CLS are different measurements and demand different repairs. A single score hides those distinctions.
| Symptom | Likely investigation | First experiment |
|---|---|---|
| Blank hero for several seconds | Image weight and resource priority | Resize the hero and preload only what is needed |
| Delayed menu or search input | Long JavaScript work | Defer unused scripts and profile input handlers |
| Cards jump while scrolling | Missing reserved dimensions | Set image width, height and aspect ratio |
| Good desktop, poor phone | CPU and network sensitivity | Reduce main-thread work and third-party tags |
What not to optimize first
Do not remove an essential form or make buttons difficult to tap in pursuit of a lab score. Also avoid adding dozens of preloads: they compete for bandwidth and can delay the actual hero resource. First fix access errors, broken journeys and severely oversized assets; then refine the slower templates. Our monthly maintenance plan includes a repeatable performance review.
