Understand the three metrics
Largest Contentful Paint measures how quickly the main content appears. Interaction to Next Paint measures responsiveness to interactions. Cumulative Layout Shift tracks unexpected movement. The good thresholds are LCP at or below 2.5 seconds, INP at or below 200 milliseconds, and CLS at or below 0.1 at the 75th percentile.
Tie each metric to what a visitor experiences
Largest Contentful Paint measures when the largest eligible visible content element renders. Interaction to Next Paint measures responsiveness to interactions over a visit. Cumulative Layout Shift reflects unexpected movement. The recommended good thresholds are LCP at most 2.5 seconds, INP at most 200 milliseconds and CLS at most 0.1. Field assessment uses the 75th percentile, looking at mobile and desktop separately.
- LoadingFind when the primary content becomes visible
- InteractionTrace work that blocks an input from rendering
- StabilityIdentify elements that push content unexpectedly
Find the bottleneck by template
Use field data to see what real visitors experience, then reproduce issues with lab tools. A slow hero image, blocked main thread or unsized embed can point to different fixes. Check mobile and desktop separately.
Separate real-user measurements from laboratory diagnostics
When Chrome User Experience Report data is available, use it to understand how real eligible visitors experience a page or origin. Lab tests reproduce a controlled condition and help inspect what caused a problem. A fast local test can coexist with slower field data because phones, networks, traffic patterns and caches differ. Check whether a report is page-level or origin-level before picking a fix.
Verify changes under real use
Give images dimensions, reduce unnecessary script work and prioritise the resources the first screen needs. Re-test representative pages and watch field data over time. Measure form completion and usability alongside performance scores.
Choose a fix that addresses the metric and template
For LCP, investigate server response and when the main image or text can render. For INP, inspect interaction handlers, JavaScript execution and rendering work on the main thread. For CLS, reserve space for images and embeds and avoid inserting content above something a person is using. Test key journeys again after changes and allow time for rolling field data to reflect releases.
Diagnosing three symptoms on one product page
A shop delays its hero product image while loading heavy scripts. Its filters respond sluggishly after users start browsing, and the review widget pushes the buy button down. The team prioritises the image, reduces filter work and reserves widget space, then checks key tasks on representative phones and watches field trends.
Write up the test device, connection and page template before comparing results. Run the first-load experience and the interaction buyers actually use, such as opening filters or starting a booking. A page with a good initial load can still respond slowly after third-party scripts or interface components have started working. Validate the fix against the affected template and watch for regressions on the mobile devices your visitors actually use.
| Metric | Good threshold | First place to look |
|---|---|---|
| LCP | 2.5 seconds or less | Server, primary resource and render timing |
| INP | 200 milliseconds or less | Event handling and long tasks during interaction |
| CLS | 0.1 or less | Unreserved media space and late layout changes |
Common questions
Is a Lighthouse score the same as Core Web Vitals?
Lighthouse gives lab diagnostics for a simulated visit. Core Web Vitals field assessments use distribution across real visits when enough data is available.
Do good scores guarantee higher rankings?
They support a better experience. Search visibility also depends on relevance, content and other signals; use the scores to diagnose problems visitors can feel.
What if my page has no field data?
Use lab tests to find issues and inspect real-user measurements when you have suitable coverage. Treat origin-level summaries as a clue about shared patterns, not a score for every page.
What to check next
- LCPCheck the largest visible element and its loading path
- INPTrace slow interactions and main-thread work
- CLSGive media dimensions and stabilise dynamic content




