At least, so says a peer-reviewed scientific paper in a well-known top journal:
The latest Huberman Lab episode quotes and summarizes this paper, among other protocols for motivation and achieving goals:
So, if you have less than 15% failure rate, it’s probably too easy for you.
Interesting! I would usually get 80-85% on the last couple of review sessions, but in rare cases 100% if the words were easy or 70+% if the session had a lot of leeches.
But I guess that’s learning! 
Not sure how that translates to Anki, because there scoring is not 0/1.
Yeah I initially went with the 80/20 rule. If I was getting more than 80% right, go faster and/or study less. Now I’m happier with something in the 90s because it keeps the re-review workload down more.
I usually get 96%-100%, but that’s because I tend to take my time with lessons, and I review items I failed, which keeps me pretty consistent.
Plus, you can’t really choose the difficulty of wk, since all level items are locked in. I guess you could increase your number of reviews to make it harder, but at that point you’re just challenging your endurance LOL
It actually looks like great measure to estimate if I should going with new lessons or postpone them for a bit untill success rate improves.
You found it when I was wondering how should I approach taking new lessons to keep best learning pace, avoiding overloading myself. Thanks a lot for this finding!
Perhaps I’m interpreting this wrong, but in the abstract it’s mentioned that the optimal succes rate is determined in the context of a binary classification task. That does not seem particularly applicable to learning Japanese (be it kanji, grammar, or other relevant skills).
Note: I did not read the entire paper, nor watched the video, so I might be completely off base here.