K-pop members outdraw their own groups nearly four to one — and the one exception turned out to be a hole in Wikidata

Ask which is read more often, a K-pop group or the people in it, and the answer is not close. Across the 86 groups where Wikidata records at least three members and every one of them has an English Wikipedia article we can measure, the members’ pages were opened 2,013,997 times in thirty days against 519,087 for the groups — 3.88 to one. In 80 of those 86 groups the members won.

That is the finding. The more useful part of this piece is the finding we nearly published instead.

The gradient that was not there

Sorted by the year a group formed, the same measurement produces a clean and satisfying story.

Read that column and the conclusion writes itself: older groups are read through their members, newer groups are read as a single object. It is a plausible claim about how audiences arrive at a group, and it would have been the headline.

Then we put one more column beside it — how much of each era’s listed membership we could actually measure.

The thing that falls across the eras is not only attention. It is coverage. A newly debuted act’s members are the least likely people in K-pop to have an English Wikipedia article, and a member with no article registers as zero views — indistinguishable, in the arithmetic, from a member nobody looked up.

Restrict the comparison to groups whose entire listed membership is measurable, and the gradient goes away:

Members outdraw the group in every era. The high point is 2010–2014, not the 1990s. The last two rows hold five groups and two groups, which is not enough to say anything about recent debuts at all — and saying nothing is the correct outcome there, not a smaller version of the original claim.

Fromis 9

The clearest way to see the hole is a group whose name contains its own headcount.

Fromis 9 debuted with nine members. Its article was opened 18,953 times in the thirty days. Wikidata records two members for it.

It is not alone. Babymonster’s group page drew 37,936 views — fourteenth of the 816 groups in the panel — and Wikidata lists two members for it. Meovv: 21,200 views, two members listed. KiiiKiii: 9,287 views, one. The Boyz: 9,213 views, two.

None of that is an error in Wikidata. It is a volunteer database that records what somebody has got round to recording, and newly debuted members are the last to be written up. But if you take P527 as a membership list and divide, you will conclude that fans of new groups are indifferent to the individuals — when what you have measured is that nobody has written the pages yet.

Across all 816 groups in the panel, Wikidata records membership for 263 of them. Two-thirds of K-pop groups have no membership recorded at all. That is the real coverage figure, and it is the first number anyone building on this data should be told.

What survives

With the artefact removed, three things stand.

Members outdraw their group, and by a lot. 3.88 to one where the comparison is clean, 2.47 to one if you include every group with at least one measurable member. In 93% of the clean cases the members won. Girls’ Generation’s nine members drew 130,642 views against the group’s 39,036. Apink’s seven drew 94,840 against 12,776 — 7.4 to one, for a group formed in 2006.

The biggest groups are the least lopsided. BTS drew 383,622 views to its group page, more than any other act in the panel, and its six measurable members drew 512,591 — 1.34 to one, against a panel norm near four. NewJeans sits at 1.06. A group large enough to be a household name in English is looked up as itself; the ratio is a rough measure of how far a name travels on its own.

Where the group does outdraw its members, it is worth asking whether the members exist in the data at all before concluding anything about fans. That is the whole lesson of the middle section, and it applies to any panel assembled from a volunteer database — which, for K-pop measured from outside Korea, is most of them.

What this is not

Every number here counts openings of an English Wikipedia article. It is not streams, not sales, not chart position, and not attention inside Korea. It says how often somebody looked a name up in English, over thirty days, which is a narrow thing to know and the only per-artist demand signal published openly and daily for everyone in this panel.

Thirty days is also all we have. We began collecting daily this month. A month from now this article could be written about trends; today it cannot.