K Culture WireKorean pop culture, in numbers

SM, YG, JYP, Hybe: how many artists we can actually count

Name: SM, YG, JYP, Hybe: how many artists we can actually count
Creator: K Culture Wire

We hold a record label for 1,044 Korean entertainers, and an English Wikipedia reader count for 650 of them. Grouping labels into the listed parent company gives 9 companies with at least 20 artists we can measure.

Do not read this table as a ranking. The share of each company’s artists we can measure runs from 37.2% to 66.7% — a spread of 1.8x. The medians run from 66 to 208 a day — a spread of 3.2x. So coverage does not explain the whole order, but a 1.8x difference in what we can see is far too large to read the medians as if every company were measured equally.

Hybe is the clearest case: we can measure only 37.2% of the artists we know it holds (35 of 94). Read its median as the median of that half, not of the company.

What we counted

What we counted	Number
Korean entertainers we hold	9,249
Of those, with a record label recorded on Wikidata	1,044
Company–artist places counted (an artist on two companies counts twice)	1,297
Of those places, with a reader count we could join	650
Coverage across everything	50.1%
Companies in the data	156
Companies with at least 20 measurable artists	9

Every company we can measure

Ordered by how many artists we could measure, largest first. Not by the median — the order a table is printed in is itself a claim, and we are not making that claim.

Company	Ticker	Labels	Largest single label	Artists we know	Of those, measured	Coverage	Median daily readers
YG Entertainment	122870	2	101	109	57	52.3%	198
SM Entertainment	041510	1	133	133	55	41.4%	102
CJ ENM Entertainment Division	130960	4	38	69	36	52.2%	66
Kakao Entertainment	—	5	42	67	36	53.7%	80
Hybe	352820	7	44	94	35	37.2%	141
JYP Entertainment	035900	2	70	72	33	45.8%	208
FNC Entertainment	173940	1	48	48	32	66.7%	106
Cube Entertainment	182360	1	38	38	25	65.8%	78
DSP Media	—	1	46	46	24	52.2%	88

147 more companies appear in the data with fewer than 20 measurable artists. They are counted, not hidden — we simply do not print a median for them, because with a handful of artists one well-known name sets the whole figure.

Why the labels are grouped

A listed K-pop company usually owns several labels, and Wikidata records the label, not the parent. Count labels separately and a company with many of them looks small. The two columns above let you see that for every row: compare “largest single label” against “artists we know”. Where those two numbers are far apart, counting by label would have understated the company.

Every figure above is grouped to the company that is actually listed on an exchange. Where Wikidata records an owner who is a person rather than a company, we keep the label as its own entity — a person does not have a ticker, and that is how we tell the two apart rather than deciding it ourselves.

What this does not say

It is not a roster. We counted only artists for whom Wikidata records a label. 1,044 people out of the 9,249 we hold — so every artist count here is a floor.
The measured half is not a random half. Reader counts come from rosters built around charting titles and most-read lists, so the artists we can measure skew towards the better-known. Every median here is therefore pulled upwards, and pulled upwards by different amounts for different companies.
Readers are not revenue and not quality. A reader count says what people looked up last month. It says nothing about records sold, concerts played, or how a company is run.
A record label is not a management contract. Wikidata’s field is the label that released the music, which in Korea usually but not always matches the agency.
An artist on two companies’ labels is counted in both, so the artist columns add up to more than the number of people.

What we wrote from this data

Five agencies pushed releases on the same day. We can measure the reading for nine labels — and 147 we cannot
Cube, SM and Hybe all had something out this week. For each of nine K-pop companies we hold a reader count, but the share of the roster we can see runs 37% to 67% — so the ranking everyone wants is the one thing the data will not give.
8 September 2026