The decision that matters
Use field data where available and controlled tests to investigate specific issues. Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift describe different aspects of experience. Avoid reducing them to one decorative score. Inspect which resource or task causes the observed delay, then make a focused change. A smaller image may help loading while doing little for a long-running interaction handler.
A practical scenario
A marketing homepage may load a large hero image late, while a dashboard responds slowly because filtering blocks the main thread. Compressing the dashboard logo will not fix the filtering delay. Conversely, rewriting an event handler will not correct a hero image without reserved dimensions that shifts the page as it appears.
Before you ship
- Identify the affected metric and representative user conditions.
- Reserve space for images and other late-loading content.
- Measure the same scenario before and after the change.
A useful next step
Keep a short performance log with the bottleneck, change and observed result. Test both desktop and constrained mobile conditions. Treat laboratory scores as diagnostic evidence and field measurements as a separate view of the experience people actually receive.
Primary reference & further reading
This guide combines general engineering recommendations with an illustrative scenario. Provider behavior can change; check current documentation before applying it. Read our editorial approach.