hiraeth node.studio
Knowledge
ENDE

Home / Guides / Measure load time

Guide · PageSpeed Insights, in the browser, no account

Measure how fast your site really is on a phone

On your own machine over wifi every site is fast. Your customers arrive on a phone and often on mobile data, and that is where it is decided whether someone waits or goes back. PageSpeed measures exactly that case and costs nothing.

Why this matters is explained in Core Web Vitals explained: LCP, INP and CLS without the jargon.

How to do it

  1. Open pagespeed.web.dev. If a cookie notice appears, dismiss it with „Ok, Got it.“, otherwise the field below will not accept anything.That is exactly what defeated our first attempt: the tool answered „Enter a valid URL“ while the address was sitting in the field.
  2. Put your address into the field „Enter a web page URL“, with https:// in front, and click „Analyze“. Use the page most people land on first, usually the home page.
  3. Wait. The report takes between thirty seconds and a good minute. „Report from“ with a date and time appears at the top when it is done.
  4. Above the result sit two tabs, „Mobile“ and „Desktop“. „Mobile“ is preselected, and that is the number that counts. The desktop score is nearly always flattering and says little about your customers.
  5. The section „Discover what your real users are experiencing“ will say „No Data“ on smaller sites. That is not a fault and not bad news: it holds measurements from actual visitors, and Google needs more traffic for that than a local business has.Our own site says „No Data“ there as well. Anyone who reads that as a problem goes looking in the wrong place.
  6. Below, under „Diagnose performance issues“, sit the circles: „Performance“, „Accessibility“, „Best Practices“ and „SEO“. Only the first one is about load time. Green from 90, amber from 50, red below that.There is now a fifth, „Agentic Browsing“. It is new and we base no advice on it yet.
  7. Keep scrolling. Below the big circle it lists the individual items costing you the most, each with the time it would save. Work down from the top, that order is the order of effect.

How you can tell it worked

Run the same test twice in a row. The number drifts by a few points because it is simulated in a data centre rather than measured on your device. Only a difference beyond roughly ten points is a real change; anything smaller is noise.

The number is a lab measurement, not your visitors' experience. It is good for comparing the same page before and after, and poor as a grade. A 100 in „SEO“ only means a handful of technical boxes are ticked and says nothing about whether anyone finds you.