20k support employees sound eminently reasonable for a billion users. If not a bit low. Having worked a bit with support I think 10 minutes is a gross overestim

20k support employees sound eminently reasonable for a billion users. If not a bit low. Having worked a bit with support I think 10 minutes is a gross overestimate for average issue resolution time, perhaps even an order of magnitude off. However, an analysis like this is probably best served by a probably distribution of solve times (50% of issues take 1 minute, 20% of issues take 5 minutes...) instead of a point estimate, and that’s where things have the potential to go off the rails. Your really gnarly support issues are (a) where the customer can’t articulate the problem coherently, (b) where there is undefined/unexpected product behavior or (c) both. A is astonishingly common and unfortunately b isn’t too rare either (bugs, undocumented features, ill communicated launches). A seems to be essentially intractable (teaching your customers how to communicate better is probably the responsibility of the public school system) but I’ve seen some novel attempts at solutions. I’ve personally used a tool called FullStory to replay a customer session to understand what’s going on but that solution probably doesn’t scale well and there may be privacy concerns based on the product area. B isn’t a tech problem at all. It is best solved by better internal processes around catching bugs and communicating product changes to users.