Web Performance: The Metrics That Really Matter
LCP, FID, CLS... Beyond the acronyms, how to concretely optimize your site's speed and its SEO impact.
Google has rendered its verdict: performance is a ranking criterion. But among the dozens of available metrics, which ones truly deserve your attention?
Core Web Vitals: The Essential Trio
Google has simplified things by defining three priority metrics:
LCP (Largest Contentful Paint)
What it measures: the time before the largest visible element is displayed.
Target: < 2.5 seconds
The usual culprits:
- Unoptimized images
- Fonts that block rendering
- JavaScript that delays loading
Quick fixes:
<!-- Preload the main image -->
<link rel="preload" as="image" href="/hero-image.webp">
<!-- Fonts in display swap -->
<link rel="preload" as="font" href="/font.woff2" crossorigin>
FID (First Input Delay)
What it measures: the delay between the first user interaction and the browser's response.
Target: < 100 milliseconds
The usual culprits:
- Heavy JavaScript blocking the main thread
- React hydration taking too long
- Third-party scripts (analytics, chat, ads)
Quick fixes:
- Aggressive code splitting
- Lazy loading non-critical components
- Defer/async on non-essential scripts
CLS (Cumulative Layout Shift)
What it measures: the visual stability of the page (elements that "jump").
Target: < 0.1
The usual culprits:
- Images without dimensions
- Dynamically injected ads
- Fonts that change size on load
Quick fixes:
<!-- Always define width and height -->
<img src="photo.jpg" width="800" height="600" alt="...">
<!-- Reserve space for ads -->
<div style="min-height: 250px;">
<!-- Ad loaded here -->
</div>
Beyond Core Web Vitals
TTFB (Time To First Byte)
The time before receiving the first byte from the server. If your TTFB is > 600ms, no frontend optimization will compensate.
How to improve it:
- CDN to bring content closer to users
- Server cache (Redis, Varnish)
- Database query optimization
Total Blocking Time
The total time the main thread is blocked. It's often JavaScript that's the problem.
Diagnosis: Chrome DevTools > Performance > Main thread
Case Study: E-commerce Site Audit
Here are the optimizations applied to a real site:
| Metric | Before | After | Improvement | |--------|--------|-------|-------------| | LCP | 4.2s | 1.8s | -57% | | FID | 180ms | 45ms | -75% | | CLS | 0.25 | 0.05 | -80% |
What was done:
- Converted images to WebP with lazy loading
- Removed 3 unnecessary third-party scripts
- Switched to Next.js for SSR
- Set up a CDN (Cloudflare)
Business impact: +23% conversion rate, -15% bounce rate.
Essential Tools
For Measuring
- PageSpeed Insights: the Google reference
- WebPageTest: detailed tests with network waterfall
- Chrome DevTools Lighthouse: local audits
For Continuous Monitoring
- Google Search Console: real data from your users
- Vercel Analytics: if you're on Vercel
- SpeedCurve: advanced monitoring (paid)
The Mistake to Avoid
Don't optimize for benchmarks at the expense of UX. I've seen sites with a Lighthouse score of 100 but a mediocre user experience (choppy animations, slow interactions).
Metrics are a guide, not an end in themselves. The final goal always remains: the user finds what they're looking for quickly and pleasantly.
Where to Start?
- Measure your current situation (PageSpeed Insights)
- Identify the main problem (often LCP)
- Apply quick wins (images, fonts, scripts)
- Measure again
- Iterate
Performance is a marathon, not a sprint. Regular improvements are better than a massive overhaul never completed.
