2

We released some measures to improve PSI score in our project.

Prior to the release, we measured Lighthouse scores in a verification environment and confirmed that they improved.

However, there was an incident where the score (including each indicator) improved in Lighthouse, but worsened in PSI.

We understand that the scores do not match, but is it possible that the trend of improving/worsening numbers differs between the before and after measures?

Lighthouse：

【Score】 70→74 (improved)
【FCP】 1.0 → 0.6 (improved)
【TTI】3.1→3.0 (improved)【SI】3.6→6.7 (worsened)
【TBT】0→140 (worsened)
【LCP】 2.1→1.6 (improved)
【CLS】 0.207 → 0.048 (improved)
　※Conducted in secret mode

PSI：

【Score】 62→44 (worsened)
【FCP】 1.1 → 0.6 (improved)
【TTI】 3.8 → 3.8 (no change)
【SI】 4.8 → 4.6 (improved)
【TBT】 60→250 (worsened)
【LCP】 2.9 → 5.6 (worsened)
【CLS】 0.026 -> 1.081 (worsened)

Incidentally, if you look at the Network tab of the Developer Tools, DOMContentLoaded and Loaded have both improved and are faster to experience.

DOMContentLoaded] 3.06s -> 1.94s Loaded] 4.97s → 3.83s

Which is correct, PSI or Lighthouse?

cursorrux
  • 1,382
  • 4
  • 9
  • 20
junichi
  • 21
  • 1
  • Which is correct, PSI or Lighthouse? Both and neither. It looks like you have some sort of race condition in the way your JS is loaded judging by the TBT times being so different. This could be something as subtle as the round trip time being higher to the PSI server than your local machine. Do you want to share a URL so I can run a test for you and then might be able to at least see which one seems more likely to be correct. – GrahamTheDev May 24 '21 at 08:18
  • Thanks for the answer, the URL is "https://theshopyohjiyamamoto.jp". I measured it with desktop. – junichi May 24 '21 at 14:27

0 Answers0