How does this improve our request desktop site experience?  The only thing this would allow us to know what sites are most often lacking functionality for users.

Wouldn't we rather know the number of times users request desktop site per session?  Per single domain over time (but scoped to a user)?  Do the characteristics of tablets differ from phones?

I don't know how rappor would give us anything other than a signal to the web dev to say their website isn't mobile optimized, but that isn't towards improving our request desktop site experience but more towards the goal of removing it (I guess we could argue that is improving it, but seems like a stretch).

I'm not necessarily against adding this, but I'm not seeing the how this will improve anything.

My thinking was about better understanding why users use "request desktop site" and whether we could do a better job of assisting them. For example, being smarter about when we present request desktop site or offering better functionality for the site (e.g. surfacing actions). Granted this will likely be a further out task, but it'd be good to have the groundwork as these sites are likely the ones where we could have the biggest improvement. Also - it isn't clear to me that we understand why people are requesting desktop site: it could be for sites that are lacking functionality in mobile or it could be for sites that don't have a great mobile layout, etc. so seeing what the sites are will help us judge better.

I agree that those other metrics are useful and we have some of the metrics available via histograms. Per single domain doesn't exist but could be informative for a sticky setting. Do you know if that's something that'll be easy to instrument?