TTI metric has a big difference between Chrome produced TTI vs Lighthouse TTI.

[abhagupta]

Trying to understand some the TTI metrics produced by private instance of WPT.. I am seeing that there is a large variance of numbers in the metrics between what is is produced by Chrome vs what is retrieved from Lighthouse metrics when lighthouse is turned on.

   FirstInteractive: 20086,
  TimeToInteractive: 56528,
  TTIMeasurementEnd: 111386,
  LastInteractive: 109465,
  step: 1,
  effectiveBps: 20265,
  effectiveBpsDoc: 442007,
  domTime: 0,
  aft: 0,
  titleTime: 0,
  domLoading: 0,
  server_rtt: 0,
  smallImageCount: 4,
  bigImageCount: 1,
  maybeCaptcha: 1,
  'lighthouse.Performance': 0.71411764705882,
  'lighthouse.Performance.first-meaningful-paint': 1274.2,
  'lighthouse.Performance.first-interactive': 6754.31,
  'lighthouse.Performance.consistently-interactive': 11751.014,
  'lighthouse.Performance.speed-index-metric': 3055,
  'lighthouse.Performance.estimated-input-latency': 65.5,
  'lighthouse.ProgressiveWebApp': 0.54545454545455,
  'lighthouse.Accessibility': 0.91954022988506,
  'lighthouse.BestPractices': 0.75,
  'lighthouse.SEO': 0.77777777777778,

Notice the difference between FirstInteractive: 20086 and lighthouse.Performance.first-interactive: 6754.31 . I am wondering why?

FYI, this is a scripted test with around 3 steps. I could also reproduce this in public instance of WPT. https://www.webpagetest.org/result/180727_RJ_84cdfa1deb3bb0430a309f62ac83b107/

The First Interactive on test ui = 27.455s while the lighthouse.Performance.interactive has value of 10856.382 in the Raw test results. Could it because of scripted test?

[pmeenan]

You can't do a multi-step script with lighthouse. It can test a single URL and that's it. The two are measuring completely different experiences.

[abhagupta]

Thanks Patrick.. My test script has a logData 1 right before the page I wanted to measure.. What did the 2 metrics measure in this scenario and which one should we actually capture..

[tkadlec]

We're in the process of cleaning up issues on this project in order to ensure we're able to stay on top of high priority bugs and feature requests. As a part of this process, we're automatically closing any issues that have had no activity within the last two years, including this one, since the codebase has changed dramatically in that time. If you feel this is still relevant, please file a new issue using the relevant issue template so we can get it prioritized. Thanks!