What a Lighthouse score actually tells you — and what it doesn't

Lighthouse runs your page on a simulated device with a simulated network and reports what it saw. That is useful. It is also not the same question as: was the site fast for the people who actually visited it last month?

Lab data versus field data

Lab data is one run, on one simulated machine, under conditions you chose. Field data is collected from real visits on real devices and networks. They answer different questions, and a site can score well in the lab while real users on older phones have a much worse time.

The numbers worth watching

• Largest Contentful Paint — when the main thing finished rendering.
• Interaction to Next Paint — how quickly the page responds when touched.
• Cumulative Layout Shift — how much the page moved under the reader.

The composite score is a weighted blend of those and more. It is a convenient summary, but the individual numbers are what you fix. A single score cannot tell you whether the problem is a heavy image, a blocking script, or a layout that jumps when a font arrives.

What to ask whoever built your site

• Can I see the field data, not only the lab score?
• Was the Arabic page tested separately from the English one?
• Which specific metric is weakest, and what is the plan for it?
• What is the performance budget, and what happens when a change exceeds it?