Connect metrics with page behaviour
Check main-content loading, responsiveness to interaction and layout stability. Core Web Vitals describe these through LCP, INP and CLS. Field data represents real visits; lab data provides reproducible tests. A high score from one tool does not prove that every visitor receives a fast experience.
Start with the largest resource costs
Choose image dimensions and formats for their actual display size, and specify width and height. Do not unnecessarily defer the primary image; lazy-load secondary media. Avoid loading video players, maps and chats before they are needed. Every external script adds a dependency, so record its purpose and owner.
Use comparable scenarios for changes
Select representative pages such as home, catalogue, article and enquiry form. Compare the same conditions before and after changes, measuring task completion as well as bytes. Do not remove useful features solely to improve a score. Agree resource limits and acceptable regressions as part of acceptance, then check them during updates.
- Images have deliberate sizes and formats.
- Third-party scripts have a stated purpose.
- Tests cover different page tasks.
- Metrics are compared with real experience.
Project questions
Can a maximum speed score be guaranteed?
Not without knowing the content and testing conditions. Agree concrete metrics, devices and scenarios, then monitor regressions instead.