Most K-pop attention is flat. At the very top it is not — a quarter of BTS's month happened on one day

Somebody selling you an attention figure owes you one more thing beside the number: whether it is a level or an event. A name that draws 60,000 views a month steadily and a name that draws 60,000 because of one Tuesday are the same row in a spreadsheet and completely different facts.

We can now answer that for 1,698 Korean music acts. For each one, take the largest single day of the thirty and divide by the month.

Under a perfectly even month that share would be 3.33%. The median act sits at 5.5% — close to flat, a little lumpy, the shape you would expect from ordinary weekday-and-weekend traffic.

The lumpiness lives at the top

The bottom half and the top hundred are indistinguishable. The pattern only appears in the top fifty and sharpens hard in the top twenty, where 11 of the 20 largest names put at least a tenth of their month into one day.

This is the reverse of what a spreadsheet intuition suggests. Small numbers are the ones that usually look jumpy, and here they are the calm ones. Whatever produces a spike — an announcement, a release, a piece of news — appears to reach the largest names and stop well short of the rest.

BTS: 25.1% of a month in one day

BTS drew 383,622 views over the thirty days, more than any other act. 96,360 of them — 25.1% — landed on 2026-07-19.

It was not the group page alone. Six of BTS’s seven members have English Wikipedia articles we can measure, and all six peaked on that same date:

Seven pages, one date, no exceptions. We do not know what happened — nothing in this dataset records why a page was opened, and we are not going to guess in a piece about measurement.

What we can say is how far it travelled. 40 acts across the whole panel peaked on 2026-07-19, against 57 on an average day. The date was not a big day for Korean music generally. It was a big day for one family of pages.

Blackpink, the same month, the other shape

Blackpink drew 105,378 views, second among groups, and its biggest day is 5.4% — beneath the panel median. The members do not line up on one date. Same window, same measure, a flat month.

Neither shape is better. They are answers to different questions, and a buyer who needs a stable baseline and a buyer who needs to catch moments want opposite rows.

What this measure is, and is not

It depends on the window. Thirty days is our whole history; we began collecting daily this month. A 90-day window mechanically produces smaller single-day shares for the same artist, because the denominator grows and the peak does not. The 5.5% median is a property of a 30-day month, not of an artist, and it is not comparable to a figure computed over a different span.

The edges of the window are not trustworthy. 114 acts have their peak on 2026-07-08, the first day we hold — twice the even rate of 57. That could be a real event on that date or an artefact of where we cut. From inside a 30-day box the two look identical, so we report the number and draw nothing from it.

Small acts are excluded, and they are not hidden. 663 of the 2,361 acts fall below 300 views or 25 days of data. A name with 40 views produces a wild single-day share from one ordinary afternoon, which would be arithmetic rather than attention.

And the underlying figure remains what it has been throughout: openings of an English Wikipedia article. Not streams, not sales, not chart position, and not attention inside Korea.