Blog
Core Web Vitals: how to interpret LCP, INP and CLS
Core Web Vitals are three user-experience metrics: LCP for main-content loading, INP for interaction responsiveness and CLS for visual stability. Interpreting them requires knowing where the data comes from, which pages it covers and which device it describes. An overall performance score cannot replace that context.
Metrics and good-experience thresholds
The Web Vitals documentation defines these reference thresholds, assessed at the 75th percentile of visits and separately for mobile and desktop:
| Metric | Threshold considered good |
|---|---|
| LCP, Largest Contentful Paint | Up to 2.5 seconds |
| INP, Interaction to Next Paint | Up to 200 milliseconds |
| CLS, Cumulative Layout Shift | Up to 0.1, without a time unit |
The 75th percentile marks a point at or below which approximately 75% of observations fall. It is not the average visit. When sufficient data exists for all three metrics, each must meet its threshold for a good overall assessment. INP is the current responsiveness metric; an explanation still listing FID as part of the set needs updating.
Example: good loading, slow interaction
Imagine a mobile field report showing LCP of 2.4 seconds, INP of 280 milliseconds and CLS of 0.08. These are teaching figures, not measurements of HeyMark. Loading and stability fall within their good thresholds. INP exceeds 200 milliseconds, so the set does not pass all three.
The next task should investigate responsiveness. “Make every image smaller” does not identify what delays the response to a button press. The team needs to reproduce interactions, examine browser work and establish what blocks the response.
Keep those mobile figures separate from a favorable desktop report. They describe different experiences. Preserve the device and reporting period alongside each result.
Why PageSpeed and your test may differ
PageSpeed Insights combines Chrome UX Report field information with a Lighthouse laboratory run. Field data summarizes real experiences over a recent window; a lab run tests a load under controlled conditions. A fresh run may reflect a change before it is represented across the full field window.
Check whether the field data covers the requested URL or whether the tool shows origin data because the page lacks a sufficient sample. An origin groups the same protocol, host and port; its result does not establish the experience of every individual URL. No field data means there is no field assessment. PSI has an exception for insufficient INP data: it can pass the assessment when LCP and CLS have sufficient data and good 75th percentiles. That does not supply a good INP measurement. Insufficient LCP or CLS data prevents an overall assessment.
A load test without interactions cannot, by itself, measure the INP people experience. Lighthouse can show indicators such as Total Blocking Time that help investigate blocking work, but those values should not be relabelled as INP. Understanding field and laboratory tools prevents comparisons between values answering different questions.
Request a fix that can be verified
Send the team the URL, device, period, data source, URL-or-origin scope and problematic metric. For the example, add interactions that feel delayed and a reproducible test if available. After a fix, repeat the local diagnosis and observe field changes over the relevant reporting window.
Google uses Core Web Vitals in ranking systems, while explaining that good page experience does not guarantee top rankings. A technical improvement may make booking or reading more comfortable; establishing its effect on traffic or bookings requires measuring those outcomes separately.
Plan your next post in HeyMark.
Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.