Core Web Vitals for Indian SMB websites: a 90-day plan

Core Web Vitals are the three metrics Google uses to judge user experience on every page: Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift. They are an official ranking factor. Failing them on mobile is the single biggest reason Indian SMB websites rank below where they deserve.

This is a 90-day plan that fits a single-engineer team.

The thresholds

• LCP under 2.5 seconds (75th percentile of real users)
• INP under 200 milliseconds (75th percentile)
• CLS under 0.1 (75th percentile)

If you are above any of these, you are failing Core Web Vitals. Most Indian SMB sites fail all three.

How to measure

Use real user data, not lab data:

1. CrUX (Chrome User Experience Report): free, no setup, shows your 75th percentile over 28 days.
2. Real User Monitoring: deploy the web-vitals library and send results to your analytics.
3. PageSpeed Insights: pulls CrUX for your domain plus Lighthouse lab data.

Lighthouse and WebPageTest are for debugging, not judgment. They run on fast connections with desktop CPU. They will under-report what Indian users on cheap Android phones see.

The five biggest fixes

### 1. Image optimization

Most Indian SMB sites have 2 to 5 MB hero images in JPEG. The fix:

• Convert to AVIF or WebP (50 to 70 percent smaller at same quality).
• Use responsive sizes with srcset.
• Lazy-load below-the-fold images.
• Preload the LCP image with fetchpriority high.

Typical win: LCP from 4 to 5 seconds to under 2.5 seconds.

### 2. JavaScript bundle size

Most Indian SMB sites ship 1 to 3 MB of JavaScript on first load. The fix:

• Use the Next.js compiler with code splitting.
• Lazy-load below-the-fold components.
• Replace heavy libraries with smaller alternatives.
• Audit third-party tags.

Typical win: INP from 400ms to under 200ms.

### 3. Font loading

Web fonts often cause CLS and slow LCP. The fix:

• Self-host fonts with font-display swap.
• Preload the fonts used above the fold.
• Use font-display optional for replaceable fonts.

Typical win: CLS from 0.2 to 0.05.

### 4. Third-party tags

Chat widgets, analytics and tag managers can add 500 KB to 1.5 MB of JavaScript. The fix:

• Audit each tag. Remove tags no one looks at.
• Load tag managers with defer and gate behind consent.
• Use a Consent Management Platform.

Typical win: LCP from 3.5s to 2.5s, INP from 350ms to 200ms.

### 5. Caching headers

Most Indian SMB sites miss basic caching. The fix:

• Set Cache-Control public immutable for hashed assets.
• Set Cache-Control public for HTML.
• Enable Brotli compression.

Typical win: repeat-visit LCP from 3s to 1.5s.

Indian-specific issues

• Low-bandwidth 3G and 4G users. Default page weight under 1.5 MB.
• Cheap Android devices. Test on a 4x CPU throttle profile.
• Regional CDN coverage. Use Cloudflare, BunnyCDN or AWS CloudFront with a Mumbai edge.
• Variable power. Optimize for the worst case, not the best.

The 90-day plan

Weeks 1 to 2: Measurement. - Set up RUM with the web-vitals library. - Pull CrUX for your top 10 pages. - Build a baseline dashboard.

Weeks 3 to 8: Quick wins. - Image optimization across the site. - Remove unused third-party tags. - Self-host fonts with font-display swap. - Add caching headers.

Weeks 9 to 12: Structural fixes. - Code splitting and lazy loading. - Replace heavy libraries. - Move to a CDN with regional edge. - Switch to a modern image format.

Weeks 13 to 14: Polish. - Inline critical CSS. - Preload LCP image and font. - Run a Lighthouse CI check on every PR.

After 90 days, you should be in the green on all three metrics for 75 percent of real users.

What to skip

Skip the temptation to use a performance plugin that auto-optimizes everything. They hide the problem rather than fix it. The real wins are structural.

Also skip AMP. AMP is no longer a ranking factor, and the cost of maintaining an AMP version is not worth it for most Indian SMBs.