Two shows from one Korean company on the same chart cost each other half a place

A country’s Netflix top 10 has ten places in it. A Korean company with two series charting at once is occupying two of them, and the fear inside a distribution meeting is obvious: our own shows are taking each other’s oxygen, so hold one back.

That fear is testable. Netflix publishes each country’s weekly chart with a rank and its own “weeks in top 10” counter; Wikidata records which company produced, first broadcast or distributed each Korean title. Put the two together and you can watch a single series in a single country and ask what happened on the weeks a stablemate was beside it.

The answer is half a place.

Rank 1 is the top of the chart, so a bigger number is a worse position: a series sits lower when its stablemate is beside it. The direction is the one schedulers fear. The size is half of one rank out of ten.

The control that matters is age, not company

The obvious way to get this wrong is to compare a company’s crowded weeks against its uncrowded ones and forget that they are not the same weeks.

A title’s rank falls as its run goes on — it opens high and slides. A company’s second title usually arrives after the first has been out a while. So “weeks with a sibling” and “weeks later in the run” are largely the same weeks, and a naive comparison would hand the slide to the sibling and call it cannibalisation.

Netflix supplies the fix in its own data: a weeks-in-top-10 column. Every comparison here is between weeks at a comparable age of the same title in the same country. Split by age band, all four point the same way:

No band reaches a whole place. The largest gap is in weeks 5–8, and the smallest is in weeks 3–4, and we are not going to build a story on the difference between 0.13 and 0.76 across a few hundred pairs — the honest summary is that all four are small and all four have the same sign.

The version of this that was wrong

We started by matching weeks at exactly the same age: week 4 against week 4 of the same title in the same country. That produced 205 pairs, and it looked like a small sample.

It was not a small sample. It was a strange one. A title does not reach week 4 twice in one country, so an exact-age match is impossible by construction — the only pairs that survived were runs where Netflix’s week counter had reset, which is not a random subset of anything. Matching inside age bands instead gives 803 pairs across 648 title-and-country runs, at the cost of allowing up to a week of slack within a band. That is the trade we made, and it is the reason the numbers above exist at all.

What was thrown away

Far more was discarded than kept, and that is not a footnote. A week only counts if the same series, in the same country, at a comparable age, also spent a week without a stablemate. Most runs never overlap a stablemate at all, and a run that is always crowded contributes nothing either. What survives is the set of titles that experienced both conditions — which is the only set that can answer the question, and is not the whole catalogue.

What this cannot tell you

It cannot tell you whether a sibling took the other title’s place or whether both rose together and one simply had to sit below the other. Netflix publishes a rank, not the viewing behind it, and never publishes what was not watched. A half-place gap is consistent with mild competition for attention and equally consistent with a company that releases its weaker titles alongside its stronger ones — nothing here separates those.

Why we bothered

Because “don’t release two at once” is a rule of thumb that costs money when it is followed too hard, and nobody had put a number on it. The number is half a place. That is real, it is in the direction the rule assumes, and it is small enough that a company delaying a launch a month to avoid it should want a second reason.

The full tables — all four age bands, the discard counts, and the failed exact-age design set beside the one we used: what happens when stablemates share a chart.