Measure before changing
Capture representative mobile and desktop journeys, including loading, scrolling and interactive states. Lab tests help reproduce problems; field measurements show how actual visits behave. Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift describe different aspects of experience. A single synthetic score is not evidence that every user or route performs well.
Prioritize content delivery
Identify the element that dominates initial loading. Use appropriate image dimensions and responsive sources, reserve space for media and avoid lazy-loading the critical image. Review font files and preload only assets that are truly needed immediately. Cache rules should let stable media load efficiently while ensuring changed scripts and pages can update reliably.
Reduce interaction cost
Inspect long tasks, repeated layout calculations and work triggered on every scroll event. Use passive listeners where appropriate and defer nonessential features. Complex motion should pause when offscreen and respect reduced-motion preferences. Reducing JavaScript is useful only if navigation, forms and content remain functional in the resulting experience.
Verify improvements and constraints
Compare repeatable measurements with the baseline and test the actual production delivery path. Check layout and interactions after compression, code splitting or cache changes. Third-party tools and large media can dominate the remaining cost; identify their business value before removing them. Bring the slow routes and affected device examples so the work targets user-visible delays rather than a score alone.