[jakobegger]

> an informed decision on which to use for a particular problem

I doubt that really is the case. I think most decisions are made based either on anecdotes, or whatever someone happens to have experience with.

It's rare that you can accurately predict the kind of workloads you'll have to deal with ahead of time, and it's even rarer that the people making the decision have experience with multiple completely different stacks.

And I don't really think it matters that much. Some people solve the problem with distributed document stores and key value stores, other people use a big transactional database and just keep putting extra RAM sticks in their server... I don't think there's always an obviously "better" choice.