Case Study: Dropping an E-commerce LCP from 3.8s to 0.9s

Starting Point

Client: women''s fashion e-commerce, 180k monthly unique visitors, 3,200 SKUs. Problem: "Mobile is slow, Google rankings dropped, cart abandonment climbed to 74%."

Initial Measurements

Stack: PHP 7.4 + WooCommerce, 50+ plugins, shared hosting, jQuery-heavy theme.

Methodology: Measure First, Change Second

1. Real-user monitoring (RUM) with web-vitals.js
2. WebPageTest "Competitive Benchmark" vs 3 competitors
3. Chrome DevTools Performance profiles on home, category, product, cart, checkout
4. Wrote hypotheses linking each metric to user behavior

11-Step Optimization

1. Hosting Upgrade (TTFB 1120 → 380 ms)

Shared → LiteSpeed VPS, PHP 8.3, OPcache + JIT. Cost: $12/mo → $45/mo.

2. Database Tuning (TTFB 380 → 260 ms)

• MySQL innodb_buffer_pool_size = 70% RAM
• Slow query log revealed 12 n+1 queries
• wp_options autoloaded rows 3,200 → 420
• Transient cleanup — deleted 18k stale rows

3. CDN + Edge Cache (TTFB 260 → 90 ms regional)

Cloudflare Pro; Page Rules set static assets 1 year, HTML 2h stale-while-revalidate.

4. Image Optimization (+600 ms LCP gain)

• All product images to WebP (originals kept as fallback)
• Responsive srcset + sizes: 400px mobile, 1600px desktop
• Native loading="lazy" with first 3 eager
• Hero image with fetchpriority="high" + preload

Avg image size 380 KB → 72 KB.

5. JavaScript Cleanup (INP 320 → 140 ms)

• Removed 14 plugins; replaced two sliders with single Swiper
• jQuery + jQuery UI 420 KB → 40 KB vanilla JS
• Third-party pixels (Facebook, GTM) moved to Partytown
• GA4 partytown + delayed load

6. CSS Optimization (+400 ms LCP gain)

• Critical CSS inlined for home, category, product
• Main stylesheet 420 KB → 180 KB (purged unused Woo components)
• font-display: swap + unicode-range for localized subset

7. Font Loading

• Self-hosted fonts (stop Google Fonts)
• Preload + preconnect tuned
• WOFF2 subset 180 KB → 42 KB

8. CLS Fixes (0.18 → 0.04)

• aspect-ratio CSS for product images
• Fixed heights on ad slots
• Skeleton + intersection observer for "Popular Products"

9. Third-Party Discipline

Old site had 24 third-party scripts. Audit reduced to 9. Remaining loaded via Partytown or defer. Chat widget delayed 3s.

10. HTTP/3 + Brotli

Enabled via Cloudflare. Brotli level 6. Avg response size -28%.

11. Architecture

• Redis Object Cache for Woo category pages
• Cart and Account cache-excluded; others full page cached
• Homepage 72 products → 24 + "load more"

Before → After

Business Results (First 60 Days)

• Organic traffic: +38%
• Cart abandonment: 74% → 62%
• Conversion: 2.1% → 3.4%
• Pages/session: 3.8 → 5.9
• Mobile conversion: +68%
• Estimated annual revenue lift: ~$120k

Lessons

1. Biggest wins come from overlooked spots: hosting + DB tuning
2. Third-party scripts are the main INP killer
3. WordPress/WooCommerce isn''t inherently slow — plugin culture is
4. Image optimization is the simplest highest-ROI step
5. Never optimize without measuring — RUM + lab together

Cost and Timeline

6 weeks; 2 senior developers + 1 DevOps. Total ~$7,500. Paid back in 60 days.

FAQs

Can I do this myself?

Some steps (image optimization, cache plugins) yes. DB tuning, critical CSS, JS splitting need specialists.

Why not migrate to Next.js?

Client didn''t want to change editorial workflow. WP + Woo, properly optimized, hits targets.

Can Shopify see similar gains?

Yes but bounded. Theme/app ecosystem is heavy; Shopify Plus allows better script control. Mid-level tuning adds 40-50 Lighthouse points.

Next Step

Free performance audit for your site — contact us.