"MongoDB's CTO disagrees with this statement arguing that nearly 90% of database installations today would benefit from being replaced with MongoDB."
I used to attend "office hours" at MongoDB's office where guests ask MongoDB employees for help. Most of my questions involved very complex aggregation queries (that would have been trivial in SQL) that even MongoDB employees could not solve.
While I waited to be helped I would listen to other people ask the same questions over and over and over: "How do I join these two collections? How can I <perform a transaction> and ensure both writes succeed/fail?"
Rather than say "MongoDB doesn't offer this functionality" and maybe advise them that this database isn't what they needed for the specific project, the engineers spent a majority of these office hours explaining they don't need schemas, transactions, relations and advised them how to hack together something that worked.
I don't think MongoDB is appropriate for 90% of the database installations out there. I don't know the number, but it isn't 90%, it isn't a majority, and doubt it's a significant number.
MongoDB really needs to explain what it does exceedingly well versus the other distinct offerings in databases (RDBMS, Key Value stores, MPP/warehouses, HDFS, etc) and market that.
What they do right now is somewhere between disingenuous and down right negligent.
Of course its the responsibility of a company and its staff to choose the right tool for the job, but these postmortems are becoming tiring.