Spaced Repetition for Kanji: How SRS Actually Works

Re-reading a kanji list feels productive and barely works — you recognize the character on the page, not in the wild. Spaced repetition (SRS) fixes this by testing recall right before you'd naturally forget, which is the single most efficient point to review.

The idea behind the algorithm

Every kanji you learn decays from memory on a curve — fast at first, slower over time. An SRS tracks that curve per card: answer correctly and the next review gets pushed further out (a day, then a week, then a month); answer wrong and the interval resets to nearly zero. Over months, cards you know well surface rarely, and cards you're shaky on surface constantly — the system automatically spends your time where it's needed.

Setting it up for kanji specifically

• One fact per card. A card that asks for reading, meaning, and stroke order at once is testing three things badly instead of one thing well.
• Learn kanji inside a word, not in isolation — a card for 食べる teaches the reading you'll actually use, where a bare 食 card teaches a reading you'll have to relearn in context anyway.
• Cap new cards at 10–15 a day. SRS backlogs compound — adding 30 new kanji daily buries you in reviews within two weeks.
• Review daily, even briefly. A missed day doesn't lose progress, but a missed week lets a lot of cards mature past their optimal review point.

The mistake almost everyone makes

Grading yourself "correct" when you recognized the answer after seeing it, rather than before, quietly breaks the whole system — the algorithm schedules the next review based on how hard recall actually was, so generous grading just means you'll see an unlearned card again in a month, having told the system you already know it.

SRS vs. an app that schedules it for you

A general-purpose flashcard app makes you build every deck by hand. A kanji-specific app can pre-load real vocabulary, stroke order, and audio per character, so the SRS engine has better material to schedule in the first place — the algorithm is only as good as the cards behind it.