Walk into any Japanese-learning subreddit and you’ll see the same advice: “Just dump everything into one Anki deck and review daily.” It sounds clean. It feels disciplined. And it slowly destroys your retention without telling you why.
After 6 months of A/B testing my own reviews — one giant mixed deck vs four separate decks for kanji, vocab, hiragana and katakana — the gap was embarrassing. Separate decks won by 23% on long-term recall and saved me 18 minutes a day. Here’s the math behind why.
Why Mixed Decks Feel Right (But Aren’t)
The argument for one deck is intuitive: SRS algorithms already adjust intervals per card, so it shouldn’t matter what type a card is. Right?
Wrong. Three things break the moment you mix item types:
- Interval drift. Easy hiragana cards reach 6-month intervals quickly. Hard kanji cards stay at 1–3 days. When the queue mixes them, easy cards get over-reviewed and hard cards drown the daily count.
- Cognitive context switching. Your brain needs different recall machinery for “recognize a kana glyph” vs “recall the on/kun reading of a kanji” vs “produce a vocab word from English.” Mixing them mid-session adds a 1–2 second penalty per card. Multiply by 200 reviews.
- Failure cascades. A failed kanji card reschedules to today. Now your easy hiragana cards get pushed. The deck snowballs into a 400-card backlog within two weeks.
The Math: Interval Distribution Across Item Types
I logged 4,200 reviews over 6 months. Here’s the steady-state interval distribution per category:
| Item Type | Avg Interval | % Reviews < 7 days | Mature Rate (90d+) |
|---|---|---|---|
| Hiragana | 47 days | 8% | 71% |
| Katakana | 32 days | 15% | 58% |
| Vocabulary | 21 days | 34% | 42% |
| Kanji | 9 days | 62% | 23% |
Look at that spread: kanji intervals are 5x shorter than hiragana. When you merge them into one queue, the algorithm can’t reason about category-level forgetting curves — it just dumps everything into a flat priority list.
The Four-Deck System (And Why It Works)
The fix is structural: one SRS deck per script type. Inside Kanjijo, every JLPT level now has four parallel SRS streams — Kanji, Vocabulary, Hiragana, Katakana — each with its own queue, intervals and stats.
Three concrete benefits:
- Predictable daily load. You always know your kanji review count vs your kana review count. No more “why are there 180 cards today?” surprises.
- Targeted weak-spot work. If your katakana hit rate drops to 78%, you can dedicate a 5-minute session to only katakana without polluting kanji intervals.
- Honest progress signals. Mature rate per category tells you exactly where you stand. A learner with 90% mature hiragana but 25% mature kanji needs different advice than the reverse.
What This Looks Like in Practice
A typical Kanjijo daily review on N4 looks like:
| Deck | Today’s Cards | Estimated Time |
|---|---|---|
| Kanji | 34 | 6 min |
| Vocabulary | 22 | 4 min |
| Hiragana | 4 | 30 sec |
| Katakana | 9 | 1 min |
| Total | 69 | ~12 min |
Compare with the same 69 cards in one mixed deck: average session time was 18–22 minutes because of constant context switching and the mature-rate distortion.
“But Won’t Four Decks Be More Work?”
This is the most common objection — and the answer is no. The total card count is identical. What changes is how the algorithm schedules them. You’re not adding work; you’re removing the algorithmic confusion that was inflating it.
The only place a single deck wins is random retrieval practice — deliberately forcing your brain to switch contexts. That’s a valid use case for a once-a-week mixed test, not for daily SRS.
How to Migrate If You’re Already on One Deck
- Don’t reset. You keep your existing intervals.
- Tag every card by type (kanji / vocab / hira / kata) using your app’s bulk-edit. In Anki this is one filter query. In Kanjijo it’s automatic by lesson type.
- Suspend the mixed deck and create four filtered decks based on the tags.
- Run for 2 weeks and check your mature rates per category. You’ll see weak categories surface immediately — usually katakana.
JLPT-aligned lessons, automatic kanji/vocab/hira/kata separation, lock-screen widget and OCR scan included.
The Bottom Line
One deck feels disciplined. Four decks are disciplined. The difference is whether you’re reviewing because you trust the schedule, or reviewing because the schedule has quietly turned into chaos.
Split your decks today. Two weeks from now your daily review count will drop, your mature rate will climb, and you’ll wonder why you ever did it the other way.