Measure first
I started from the Lighthouse results instead of guessing, and traced the score to three causes: unoptimized images, a large JavaScript bundle (whole icon packages imported at once) and render-blocking fonts.
Case study 01 — Performance
An application's Lighthouse performance went from about 55% to 95%+, and the storefronts stayed at 95%+ on every category.
( 01 ) — Context
The Sailors builds booking and e-commerce products. I was a Senior Frontend Developer there, working remotely from Palestine for a Saudi team.
The challenge
The application scored about 55% on Lighthouse performance, and new features kept landing while I worked on it.
( 02 ) — What I did
I started from the Lighthouse results instead of guessing, and traced the score to three causes: unoptimized images, a large JavaScript bundle (whole icon packages imported at once) and render-blocking fonts.
I fixed the imports, split the routes, sized the images and changed how fonts loaded: rendering optimization, code splitting, caching strategies and media and asset optimization.
Scores drift as features land. On the storefronts I checked Lighthouse before merging, not after launch, so new features couldn't quietly pull them down.
The result
Performance went from about 55% to 95%+. The e-commerce storefronts shipped at 95%+ across performance, accessibility, SEO, best practices and PWA readiness, and stayed there.
Lesson: Measure before you optimize, and make performance a merge requirement instead of a launch task.
( ✦ ) — Related
Next case study
A FIFA World Cup Qatar 2022 ticket-booking launch→