In this amusing blog post, Scott Aaronson explains why he is unwilling to deal with editorial management systems as a referee. They make you log in to some complicated system, using a password you've probably forgotten, all to provide your expertise for free. They often don't give you the sense that you are corresponding with an actual human.
And they can go hilariously wrong. In March of this year (2017) I was informed that I was supposed to have finished a referee report by May 15, 2017, that my report was therefore "-51 days late", and that I should therefore finish it as soon as possible.
From my perspective simple e-mail correspondence works better as a referee. (Many math journals make do without these systems.) However, I have never been a journal editor, and so I wonder if I am missing the point. Do editors genuinely find these systems to be more convenient for managing their workloads? Or alternatively is it publishers that prefer to use them, and thus ask the editors to go along with them?