Performance Optimization Case Study
This was not a brochure-site cleanup. The site had logged-in behavior, ecommerce-style flows, and learning-platform features that could easily break under aggressive optimization. The job was to recover performance without damaging carts, checkout, quizzes, course progression, or account-related behavior.
The starting point
Across four benchmarked pages, the average scores were 22.5 on mobile and 52 on desktop. The site was slow, but the harder part was that the usual blunt fixes were risky because too much functionality depended on early scripts and context-sensitive behavior.
What I changed
- Reviewed the stack across Cloudflare, Hummingbird, Smush, and Perfmatters instead of treating performance as one plugin setting problem
- Separated public-page optimization from sensitive logged-in, cart, checkout, and learning-related contexts
- Added scoped Delay JavaScript exclusions where early execution was required for real functionality
- Used a small MU-plugin layer to disable unsafe optimization paths only where they caused breakage
- Adjusted a single-product layout edge case and first-click mobile-menu behavior that surfaced once the frontend was optimized harder
Result
After the optimization pass, the average scores improved to 94 on mobile and 96 on desktop across the same four pages. More importantly, the site kept the behavior that actually mattered to the business instead of trading functionality for a prettier score.
Stack used
- WordPress
- Cloudflare
- Hummingbird Pro
- Smush Pro
- Perfmatters Script Manager
- MU-plugin compatibility guards
This page stays anonymized because the useful part is the engineering method: diagnose the real bottleneck, isolate risky contexts, and improve performance without breaking the site people are actually using.