Kumi data report
Twenty-five radicals appear in nine out of ten kanji
239 radicals build the 2,501 most common kanji, and they do not pull equal weight. Twenty-five of them appear in 90.7% of those characters. Ten appear in 75.6%.
Measured over the 2,501 kanji that carry a frequency ranking, because those are the ones with complete component data. The method note explains why that scope is the honest one.
How many kanji does a radical usually build?
The average radical appears in 42.2 of these kanji, which is a misleading number. The median appears in 18. When the mean sits more than twice the median, a small group at the top is carrying the distribution.
The extremes make the point. The most productive radical, 丿, appears in 599 of them. At the other end, 14 radicals appear in exactly one, across 10,090 radical-to-kanji connections in total.
Which radicals should you learn first?
The ones above, in roughly that order. Learning radicals is often presented as a complete set to memorise before kanji study begins, and the distribution argues against that: twenty-five radicals already appear in 90.7% of common kanji, and the remaining 214 mostly add depth rather than reach, many of them appearing in only a handful of characters.
The practical order is high-productivity radicals first, then the rest as they turn up in kanji you are actually learning. A radical that builds one kanji is better learned with that kanji than in advance of it.
Method
- A connection is a radical-to-kanji link in Kumi's concept graph. Distinct kanji are counted per radical, so a pair joined by more than one relation type counts once rather than inflating that radical.
- Scope is the 2,501 frequency-ranked kanji, and the 239 radicals that appear in them. This is a correctness constraint rather than a convenience one: radical decomposition exists for 6,415 of the 13,108 kanji in the dictionary, because the source data does not cover the rare tail, so a percentage computed over "all kanji" would really be a percentage of whichever half happens to have data. Every frequency-ranked kanji has radical data, so within this scope there is no gap.
- Two different measures appear here and they are not interchangeable. COVERAGE is the share of kanji a group of radicals appears in (90.7% for the top twenty-five). CONNECTION SHARE is those radicals' slice of all radical-to-kanji links (52.7%). Coverage is higher because a kanji usually has several radicals, so a character is covered if any one of them is in the group.
- Productivity is not the same as frequency in text. A radical inside many rare kanji ranks high here without being common on the page.
- Numbers are computed live from Kumi's dictionary each day. Last generated 2026-08-23.
Component and radical data derive from KanjiVG (CC BY-SA 3.0) and KANJIDIC2 (EDRDG, CC BY-SA 4.0). The productivity counts and concentration figures are Kumi's own. See attributions.
Kumi connects every radical to the kanji it builds and the vocabulary those kanji appear in, so a component is learned once and reused everywhere it turns up.