A solid foundation
Look beyond a single speed score
Core Web Vitals measure loading (LCP), responsiveness (INP) and stability (CLS). Good thresholds: 2.5 seconds, 200 milliseconds and 0.1. Assess these at the 75th percentile of visits, separately for mobile and desktop.
A controlled test simulates a situation; real visitors use different devices and networks. PageSpeed Insights can show both kinds of data when enough field data exists. Compare mobile and desktop separately and test more than just the homepage.
Work from the visible problem to its cause
Main content appearing late? Review image sizes and loading order, blocking scripts and server response. A menu or form reacting slowly? Investigate JavaScript and heavy page tasks. Content shifting? Reserve space for images and embeds in advance.
Change one important factor at a time and repeat the same test. Caching can help repeated requests, but baskets and logged-in pages need their own rules. After optimising, check that forms, payments and other dynamic parts still work properly.
Match the symptom to an investigation
| What the visitor notices | What to investigate first |
|---|---|
| Content appears late | Server response, large images and blocking resources. |
| Clicks feel slow | JavaScript, long tasks and plugin behaviour. |
| Buttons jump away | Missing dimensions, late banners and loading fonts. |
A measurable improvement cycle
- Choose a busy or important page and record a baseline.
- Note the device, testing method and specific problem.
- Address one cause and measure again under the same conditions.
- Test the user action and monitor real visitor measurements when available.
Questions about website speed
Can hosting solve every speed problem?
No. Hosting affects part of the chain. Images, themes, scripts and external services can also be the bottleneck.
Does my speed score always need to be 100?
A score is a tool. Assess whether visitors can complete important tasks smoothly and follow the relevant measurements in context.
Why do two test results differ?
Network, device, cache and background work affect results. Repeat tests and compare equivalent situations before drawing a conclusion.
Read more from the source: web.dev · Web Vitals
Bring a specific measurement
Share the page, the problem and a public test result. Mention whether it concerns mobile, desktop or a particular action.