Why are WPT results always slower than actual loads

100% of the time when I run a test with WPT (even at the fastest speeds) it is slower than the actual load times I see. Sometimes its off by maybe 20% but sometimes it can be 300% or more. I understand it’s a synthetic test and it can’t be perfect but I always have wondered this and never asked.

Also, when trying to improve CWV for SEO purposes should I be testing using one of the mobile phone emulation options (and what bandwidth) since Google does mobile-first indexing? When doing that, it’s rare for me to get all three metrics in the green zones.

Thanks

This is a tough one to say without knowing your real user data. There are a few different variables that come into play here:

  • Test location (perhaps your server or some services are slower from one geography to the next)
  • Computing power (this is a big one, and if you’re comparing WPT results to results from your own machine, a very likely candidate)
  • Network conditions (sounds like you’re maxing this out, but still a possible variable here)

If your site gets enough traffic to be in the Chrome User Experience Report (CrUX), when you test in a Chrome browser we’ll provide CrUX data so that you can compare that WPT result to what Chrome sees. Ideally you want to be testing a WPT run that is inline with your p75 data there (or a little slower) so you are catching the bottlenecks impacting your users. That can be a helpful gauge to figuring out why/where a gap occurs.

Also, when trying to improve CWV for SEO purposes should I be testing using one of the mobile phone emulation options (and what bandwidth) since Google does mobile-first indexing?

Yup! We’d recommend an emulated G4 and a 3G Fast or 4G network.

Thanks. How often are you able to get the sites you work on all green CWV using fast 4G? I’d estimate only 25% of the websites I work on would be able to get all green on fast 4G. Do I need to up my game?