French learners are obsessed with listening, Russians refuse it: a year of data on how 8 different cultures choose to study kanji
I help build a Japanese learning app. It's localized in 21 languages with 18,744 users, and I just pulled a year of numbers. One pattern keeps bugging me.
The weird design choice
Instead of 2 or 3 quiz formats like most apps, I shipped 12
🎧 Listening
🧩 Radical decomposition
✍️ Drawing
🔍 Spot the similar kanji
🐛 Find the broken one
🖼️ Emoji-to-meaning matching
🪢 Drag-and-drop
🗣️ Onyomi readings
👂 Shadowing
📝 Dictation… and more
My theory was that kanji aren't one skill, they're a bunch of small ones, and people should be free to attack from whichever angle clicks.
What I didn't expect, the angle people pick lines up almost stereotypically with their native language.
The data
📊 665,534 quiz attempts analyzed
🌍 8 native-language communities with enough volume to talk about
👥 Users behind those attempts: English (375), Italian (214), Spanish (190), French (83), Indonesian (67), Russian (45), German (43), Portuguese (28)
📐 Percentages below = share of attempts per exercise type, vs the global average
Each group's signature
🇫🇷 French, the listeners
>
🇷🇺 Russians, the structuralists
>
🇮🇩 Indonesians, the pictogram people
>
🇮🇹 Italians, the bug hunters
>
🇬🇧 English, the detail freaks
>
🇵🇹 Portuguese, softer French
>
🇩🇪 Germans, analytical but mild
>
🇪🇸 Spanish, the monoculture mystery 🤷
>
The thing that keeps bugging me
Everyone sees the same app. Same UI, same quiz menu, same catalog. Just localized text. The choice is completely free.
And yet the moment you let people pick, they cluster into these almost cartoon-stereotypical preferences. The French vs Russian polarity especially: same product, opposite ends. Listening on one side, structural analysis on the other. Almost like the phonetic-vs-orthographic intuitions baked into their native languages bleed into how they want to meet a logographic script for the first time.
Honest caveats
⚠️ This measures preference, not skill. French speakers reach for listening more, but I can't tell you they're better at it
⚠️ Some samples are small (Russian n=45, German n=43, Portuguese n=28). Big groups are solid, small ones are "indie-dev anthropology", not peer-reviewed work
⚠️ The Spanish thing genuinely stumps me. Best theory in the comments wins my eternal gratitude