The overall score of a web page in Google PageSpeed Insights cannot be directly equated with the performance of the page with the users, because different aspects go into the calculation.
At first glance, the prominently displayed score in Google PageSpeed Insights is practical. A website can score between 0 and 100 points. But how to interpret a certain number of points is not that easy to understand.
If you take a closer look, you can see in some test results that a site may score few points but still pass the Core Web Vitals test.
The cause of this discrepancy is the way Google PageSpeed Insights collects performance data. On the one hand, this includes the real user data, also called field data. They come from the Chrome User Experience Report. This is where data flows in that is transmitted directly by the browser when you visit websites.
Added to this are the laboratory data that are determined by the tool itself. These sometimes differ quite significantly from the field data. However, the laboratory data is crucial for the overall score of a website in Google PageSpeed Insights. They are based on Lighthouse Performance Scoring. Various key figures with different weightings are included here. Currently these are the following metrics (the Core Web Vitals are written in bold, the First Input Delay, also a key figure from the Core Web Vitals, is not represented here):
- First Contentful Paint: 15 percent
- Speed index: 15 percent
- Largest Contentful Paint: 25 percent
- Time to Interactive: 15 percent
- Total Blocking Time: 25 percent
- Cumulative layout shift: 5 percent.
If a website performs worse on this test, it receives a correspondingly low score. This may be due to the fact that the test found theoretical problems that are not reflected in the real user data.
John Mueller explained in the February 5 Google Search Central SEO Office Hours to use the field data to decide whether or not to improve something and to use the lab data for more detailed analysis.
It should be noted, however, that the field data is displayed with a delay of 30 days. This means that improvements are not reflected there immediately. In order to test whether a measure was successful, one should also refer to the laboratory data.
