2

I try to optimize the website https://www.onwalt.de

When I use Lighthouse within the Google Chrome DevTool, I get really good results and a time of 0.9 seconds for the Largest Contentful Paint (LCP) in desktop mode and a LCP of 2.3 s in mobile mode.

However, when I use PageSpeed Insights the LCP goes up to 8.2 seconds for mobile devices: https://developers.google.com/speed/pagespeed/insights/?url=https%3A%2F%2Fwww.onwalt.de%2F&tab=mobile

I get the same bad results with https://web.dev/measure/

Could anyone explain the difference?

I cannot find the reason. Yes, there are some render-blocking resources, but they cause max. 2 seconds.

What am I doing wrong?

The option "clear storage" is enabled: enter image description here

Phantom
  • 638
  • 1
  • 7
  • 18
  • Best guess is that you don't have "clear storage" ticked in Lighthouse on your local machine. If you don't see this on the audits page click the cog icon top right, check that "clear storage" is checked as otherwise you are comparing a warm cache (local time) to a cold cache (PSI and web.dev time). – GrahamTheDev Jan 09 '21 at 18:05
  • @GrahamRitchie Thank you very much for your input. I checked and the "clear storage" option is active. But despite the difference (local and PSI), how would you explain the 8 seconds? – Phantom Jan 09 '21 at 20:50
  • You haven't pointed your hosts file to your local host by any chance? I get similar scores to PSI when I test this. Or is this your own server and you are on a VPN to it? I get a TTFB of 570ms for nearly all requests so that is the root cause of your low scores, working out why you don't get the same is the fun part. – GrahamTheDev Jan 10 '21 at 09:38
  • @GrahamRitchie Thank you once again. No, but the server is located in Germany and so am I? More important, why is the TTFB so high? You say for nearly all request. Do you mean every request (for every file, e.g. css, img, js, etc.)? Or is it just the first request of the page (because there is a db request). – Phantom Jan 10 '21 at 11:05
  • Just my personal opinion that PSI measurements for LCP have subtly changed in recent weeks. I use https://nextjs.org/ as a reference. Usually FCP and LCP are close, but now I'm seaing numeric values like FCP 2.7s and LCP 7.6s that I wasn't seeing before (or at least not as frequently to think something problematic of it). When this started happening to my pages I initially though it might be due to subtle DOM ancestry changes after JS is execced, even though visually everything is stable, but I am not certain of such an explanation anymore. – Filip Dupanović Jan 13 '21 at 09:51

0 Answers0