What is PageSpeed Insights?
PageSpeed Insights is Google's free tool that reports how a page performs on phones and on desktops. It shows two kinds of result: field data from real Chrome users over the previous 28 days, taken from the Chrome User Experience Report, and lab data from a Lighthouse test run in a controlled setting. The field data tells you how the page feels to visitors; the lab data helps you find the cause.
How to read a PageSpeed Insights report in five minutes
Test the page that matters
Test the page buyers land on, not only the homepage: a service page, a listing or a product. Use the speed test above or Google's own PageSpeed Insights.
Read the field data first
The top section shows real users over the last 28 days at the 75th percentile. If it says the page passes Core Web Vitals, most real visits are fine, whatever the lab score says.
Check page or whole site
When a page has too few visits, the field data falls back to the whole origin, meaning every page on the site. When the site has too few as well, there is no field data and only the lab test remains.
Then read the lab score
The score out of 100 comes from Lighthouse: 90 and above is good, 50 to 89 needs improvement, below 50 is poor. It is a weighted average of five lab metrics.
Find the heaviest weight
In Lighthouse 10, Total Blocking Time carries 30% of the score, LCP and CLS 25% each, and First Contentful Paint and Speed Index 10% each. A script problem moves the score more than a slow first paint.
Work the biggest savings first
Start with the suggestions that save the most time. In the speed test above, Where the time goes lists the seconds each fix could save.
Retest, then wait for real users
The lab score changes on the next test after a fix. Field data covers 28 days, so judge the real result a few weeks after you publish.
Every number on the report, in plain words
Field metrics are judged at the 75th percentile of real visits; the lab score is a single simulated load.
| Number | What it means | Good |
|---|---|---|
| LCP, Largest Contentful Paint | When the main image or headline appears | 2.5 s or less |
| INP, Interaction to Next Paint (field) | How fast the page reacts to a tap, click or key press | 200 ms or less |
| CLS, Cumulative Layout Shift | How much the layout jumps while it loads | 0.1 or less |
| FCP, First Contentful Paint | When the first text or image appears | 1.8 s or less |
| TTFB, Time to First Byte (field, experimental) | How long the server takes to start sending the page | 0.8 s or less |
| TBT, Total Blocking Time (lab) | How long scripts keep the page from responding while it loads | Lower is better; 30% of the score |
| Speed Index (lab) | How quickly the visible part of the page fills in | Lower is better; 10% of the score |
| Performance score (lab) | The Lighthouse score out of 100 | 90 or more |
Thresholds from Google's PageSpeed Insights documentation; weights from the Lighthouse performance scoring documentation (Lighthouse 10).
Field or lab: which to trust
Trust the field data for the verdict and the lab data for the cause.
- Field data
- Real Chrome users over 28 days, across real phones and networks, at the 75th percentile. It tells you whether visitors actually have a good experience, and it updates daily.
- Lab data
- One simulated load: the mobile test emulates a mid-tier phone on a mobile network, the desktop test an emulated desktop on a wired connection. It is repeatable, so use it to test a fix.
- When they disagree
- Google says good lab data does not necessarily mean real users have a good experience. A fast lab result with poor field data usually points at real devices, networks or scripts the test does not see.
Who reads these reports
Business owners
An owner tests the page an ad sends traffic to before raising the budget, since a slow landing page wastes paid clicks.
Web developers
A developer runs the lab test before and after each change, then watches field data over the following weeks to confirm it held.
Students
A computing or design student uses the report to learn how browsers load a page, testing a project after each change and reading which metric moved.
Marketers
A marketer adds the field verdict for key landing pages to the monthly report, next to conversion rate.
Online stores
A store compares product and category pages, which usually carry the heaviest images and scripts.
IT teams
An IT team checks TTFB after a hosting or CDN change, since a slow first byte delays everything after it.
Worked examples: when the numbers disagree
A lab score of 45 that passes in the field
Suppose the lab score is 45 but the field section says the page passes Core Web Vitals. Real visitors are mostly fine; the lab test is harsher than your audience. Fix the heaviest script the report names, but do not rebuild a site to chase a lab number.
A lab score of 92 that fails in the field
Suppose the lab score is 92 but field LCP is poor. The test phone is not your buyer's phone on a busy network, and the lab run may not see the chat widget or the tag that loads for real visitors. Check the size and priority of the main image, the server first byte and every third party script, then watch the field data.
Tips for reliable readings
Test mobile first
The mobile test is the harder of the two, with slower hardware and a slower network. If mobile passes, desktop usually does.
Run it more than once
Lab results move between runs with network routing, test hardware, ads, A/B tests and browser extensions. Lighthouse advises treating performance as a range of scores, not one number.
Host near your visitors
If most visitors are in the GCC, serve the site from a nearby data centre or a CDN with a regional point of presence, then check TTFB again.
Use Search Console too
Search Console's Core Web Vitals report groups similar pages, labels each group by its poorest metric, and does so separately for mobile and desktop.
Go deeper on each metric
Our Core Web Vitals guide covers the fixes for LCP, INP and CLS one by one.
What changes in 2027 for page speed
Now
- PageSpeed Insights field data reports LCP, INP, CLS and FCP, with TTFB as an experimental metric, all at the 75th percentile of the previous 28 days.
- Lighthouse 10 weights Total Blocking Time at 30% of the performance score, LCP and CLS at 25% each, and FCP and Speed Index at 10% each.
- Google says there is no single page experience signal, and that Search shows the most relevant content even when its page experience is sub-par.
Next
- AdEngage UAE expects speed to matter more as more visits arrive from AI answers and maps on phones, where a slow page loses the visitor before it loads.
- AdEngage UAE expects more pages to fail on INP than on LCP in 2027, as chat, tracking and personalisation scripts pile onto the same pages.
- Track field data every month, and test every release in the lab before it goes live.