+benwells

I'm not quite sure I understand what you're proposing here. Is it:

X%: the percentage of page loads where the site tries to call vibrate
Y%: the percentage threshold of page loads where the site tries to call vibrate. Above this threshold, we allow vibrate on page load?

I have a few concerns:

a) it might punish users who unintentionally end up visiting a bunch of sites which all try and vibrate for bad reasons

b) it seems a little specific if we end up creating one of these per each API that we want to solve this problem for

c) this feels less like measuring user engagement, and more like implementing a per-user UseCounter

Please clarify if I've gotten any of this wrong. :)

I guess Ojan suggestion comes from an idea we had for autoplay to create a Media Engagement score where we would count how much a user uses media on a given site to unlock autoplay for them. Not sure how it can be generalised to other features.

We actually capture media as a component of site engagement right now. Time spent playing media in the foreground and background are both counted. We initially debated whether or not to store each type of engagement separately (and return a combined score), or whether to just aggregate everything together. We went with the latter approach for simplicity.

How seriously are you looking into the media engagement score?