Attributing impact on Web Vitals 🧪

TPW currently does a great job capturing the execution costs of individual third-party scripts.

I'm curious if there's a mechanism we can define for measuring the impact these third-parties have on web vitals. As an example, let's take CLS:

For each Lighthouse run via HTTP Archive/WPT:

• Check the Cumulative Layout Shift elements audit for the top N DOM elements contributing to shifts *
• Using some initiator or DOM API patching magic (wild hand waviness), discover what script initiated the injection of each of these DOM elements
• Visualize attribution in a similar way as is done for script execution

Is this crazy-talk, @patrickhulce ? :)

• Exploring this type of idea may rely on us being able to craft better 3P CLS attribution in Lighthouse itself first, so feel free to say this is not remotely possible for now. Current CLS audit looks at elements that were shifted but not necessarily why they shifted.

Start with the Lighthouse Cumulative Layout Shift elements audit and the HTTP Archive/WPT Lighthouse runs described here. Investigate whether initiator or DOM API patching can connect shifted elements to third-party scripts, then define attribution visualization comparable to the existing script-execution view. Done means a feasible measurement approach and visualization scope are established.