Funny you should mention that today. I was just reading an attempt at making sense of that: http://slatestarcodex.com/2018/11/07/ssris-an-update/

IANA doctor, but my understanding from years of trying every antidepressant on the market is that we don't currently understand why SSRIs help (some people) with depression. We know the mechanics of what they do, but not why that function has an antidepressant effect in some people. We don't really have a good idea what mechanically causes depression at all - in some cases, it's related to serotonin, in which case an SSRI might help. But there are a lot of other cases that aren't so obvious, and as far as my understanding goes, the "lot of other cases" is enough to make those who might be genuinely helped by SSRIs' mechanical function (as opposed to a placebo function, which is a whole 'nother issue in this research) be effectively invisible in the data.

A very rough metaphor (and again IANA doctor or scientist, just someone who's struggled with this personally for two decades) would be, "Stitches help people who are bleeding!" Yes, that is true in some cases and under certain circumstances, but what we've been doing with antidepressant drugs is roughly the equivalent of treating every possible bleeding wound with some variant of stitches or butterfly closures. Very effective for some, not helpful for others, actively harmful for a minority.