Home / Building & SEO / The Three Core Web Vitals Metrics, and How to Actually Bring Them Down
The Three Core Web Vitals Metrics, and How to Actually Bring Them Down
LCP, INP and CLS each have clear causes. Find which one is failing first, then optimize — don't just throw "performance plugins" at it.
"Performance optimization" is too broad a term. What you actually need to watch is three metrics with clear thresholds, each with its own causes and fixes.
LCP: Largest Contentful Paint (target < 2.5s)
LCP is usually the big image or the big headline above the fold. Causes of slowness, in order:
- The image is too large (uncompressed, not cropped to the display size)
- Fonts blocking rendering (slow-loading custom fonts)
- Slow server response (high TTFB)
- Above-the-fold content rendered by JS
Fixes, ordered by cost-effectiveness: compress the image → use system fonts or font-display: swap → add caching or change servers → only then consider changing how the page renders.
A 1200px-wide image compressed to under 100KB is entirely sufficient. This step alone usually shaves off a second or more.
INP: Interaction to Next Paint (target < 200ms)
INP reflects "how long after clicking until something happens". The usual cause is the main thread being held by long tasks:
- Loading too many scripts at once
- Heavy work inside scroll listeners
- A large batch of DOM operations crammed into a single event
Fixes: defer non-critical scripts, and split heavy work into chunks with requestAnimationFrame or setTimeout. Loading one less script is often more effective than optimizing the scripts you already have.
CLS: Cumulative Layout Shift (target < 0.1)
CLS is "the page jumping around". Three typical sources:
- Images without width/height → text gets pushed down once they load
- Ad slots inserted after the fact → content gets shoved away
- Font swap (fallback and custom font have different line heights)
The fixes are straightforward: set width/height on images, reserve a fixed height for async content, and align font metrics (size-adjust).
Measure First, Then Fix
# Run Lighthouse locally once to see the three specific scores
npx lighthouse https://your.site --only-categories=performance --preset=desktop
For mobile use --preset=mobile; the two differ a lot. Failing on mobile while passing on desktop is the norm.
One Caveat
CWV is an "experience metric", not the whole of the ranking algorithm. Sacrificing content readability for it (making text smaller, collapsing content) is a bad trade. Make sure it's readable first, then talk about scores.
How to Fix Each of the Three Metrics
| Metric | Typical cause | First thing to try |
|---|---|---|
| LCP | Hero image above the fold, font blocking | Preload the image, font-display: swap |
| INP | Long tasks on the JS main thread | Split tasks, event delegation |
| CLS | Images/ads without dimensions | Hard-code width/height, reserve ad slots |
The most common cause of CLS is "images without dimensions + fonts that load late". Fix those two and CLS is basically zero.
Field Data Beats Lab Data
Pagespeed Insights gives lab data; real-user data lives in the Core Web Vitals report in Search Console. When the two disagree, trust the field data.
A Final Note
Metrics are an outcome, not a goal. Delete every bit of interactivity to hit 100 and the site will be fast, but nobody will stay.