[robertp]

I think you have to hand it to Google, that is really next level deep analytics. I feel it would be hard for most companies without huge investment to do this (or maybe not).

It is like the reverse of a "customer survey" that instead of asking and getting an arbitrary number, instead of really detailed level of service performance.

[wora]

People often consider to use this information for reliability and performance, but you can do much more with the data. For example, if a method has low latency, you can use short deadline with fast retry to improve reliability. If you see a sudden jump of certain usage, you can consider to use batching and caching to reduce your cost. If you see an unexpected usage of a service, you know someone introduce a new dependency in your system. Google teams often use the same data to understand how large services work and how they are correlated.

Disclosure: I worked on this feature at Google.