WebVitals measures different LCP than Google Lighthouse

[SkaiiiDSO]

Hi all,

i've come across following Bug with the WebVitals-Extension

Describe the bug When Running a Lighthouse-Test on a Webpage and viewing the Web Vitals through the WebVitals Extensions the resulting LCP in Lighthouse and WebVitals-Extension differs in some Cases. EDIT: This seems to happen on mostly clientside rendered pages (for example: https://www.imwalking.de/damenschuhe/pumps/#als=true&highlight=98130727)

To Reproduce

• Open up a clientside rendered webpage
• Make sure to have the WebVitals-Extension installed
• Run a Lighthouse-Test
• After the Testrun is complete, make sure that lighthouse doesn't reload the page again. (If it does, repeat the Lighthouse-Test)
• Then Open your Web-Vitals-Extension and Compare Lighthouse Result to the Web-Vitals-Extension

Expected behavior Largest-Contentful-Paint Values should be equal in the Web-Vitals-Extension and Lighthouse-Reports

Version:

• OS w/ version: Windows 10 (19043.1110)
• Browser w/ version: Chrome 91.0.4472.164 & Chrome 92.0.4515.107

Additional context, screenshots, screencasts

[rviscomi]

Not sure this is a bug with the extension, as its CWV data is based on the web-vitals.js library. AFAIK Lighthouse uses the native APIs directly and not the wv library so that may be a source of discrepancy. cc @philipwalton @paulirish

[paulirish]

Lighthouse's LCP (as of 8.0) includes all subframes, too. (matching CrUX). As I'm guessing the extension doesn't also gather the values from subframes, this may explain the discrepancy.

If it's not a matter of subframes, this is unexpected. (I'm 85% sure our time origin definitions match up, but that'd be my next guess)

[rviscomi]

Seems like this is WAI. Closing.