A Korean title loses half its readers in two months. Most of them come back at least once

The last piece ended in a wall. We wanted to know whether a Korean hit leaves the floor higher than it found it, and for most titles that question cannot be asked: the encyclopaedia article is created when the title arrives, so there is no before to compare against.

There is a version of the question that needs no before. After the peak, how long until half of it is gone?

Of the 59 titles we hold, 26 have a peak thick enough to time and at least six months of watching afterwards. Their median half-life is two months.

How long half of it lasts

Reads are per million reads of that Wikipedia, summed across the Indonesian, Vietnamese, Thai and Malay editions.

Nine titles lost half their readers in a single month. One, Big Mouth, has been watched for 26 months and has not fallen below half — the only one of the 26 that has held.

The biggest peaks are in the top rows. Queen of Tears and Squid Game are the two largest waves in the dataset and both are gone by half within thirty days. Goblin, at a tenth of Squid Game’s height, took seven months. Whatever makes attention last, it is not size.

We report the median rather than the mean. The mean is 2.6 months, which is longer than fifteen of the twenty-six titles actually lasted.

A half-life is not an ending

Nineteen of the twenty-six later rose back above half their peak.

The eight that came back most often:

Squid Game is not on that list. It came back once, thirty-nine months after its peak, which was season two. That is the shape a release schedule makes: one return, on a long delay, when something new attaches to the name.

The eight above do not have that shape. The Handmaiden came back seven separate times, the first only three months after its peak, with nothing new attached to it. Neither does Hotel del Luna at six, or Itaewon Class at five.

The half-life counts the first fall, not the last. Read together, the two tables describe something other than a wave receding. They describe a wave that recurs: briefly, repeatedly, and mostly not on any schedule a release calendar would predict.

The thirty-three we could not time

The last row deserves care. Seven titles peaked too recently to have fallen below half yet. That is not endurance; it is the calendar. Putting them in the table above would have pulled the median up for no reason at all.

What this cannot tell you

Monthly data puts a floor under every figure here. A title that lost half its readers in ten days and one that took thirty both read as one month. The short end of that table is shorter than it looks.

The titles were chosen as ones we expected to have had a wave, not the most-read Korean titles and not a random sample. A set chosen that way leans toward titles that travelled far enough to have a peak worth measuring at all.

A read is not a viewer. Someone opening an encyclopaedia article may have watched the show, may be deciding whether to, or may have seen the name somewhere and been curious.

And we cannot tell you why Goblin held for seven months while Queen of Tears halved in one. We can tell you it was not the size of the wave.

The full tables, including every title we could not time and why, are on half-life.