Undercover took 48 chart places in our Korean panel. It is a Belgian series

A Netflix country chart gives you a title and a rank. It does not give you a work. If two productions carry the same English name, the row is the same either way, and any list built by matching title text will quietly collect both.

Ours did. Thirteen titles in our Korean panel were foreign works. The largest of them, Undercover, held 48 chart places across 24 countries. It is the Belgian series.

What gave them away

Not the name — the names are identical. Where they charted.

Undercover never appeared on Korea’s own chart. Its heaviest markets are the Netherlands and Belgium, which is where a Flemish-language Netflix series would be heaviest. Nine of the thirteen charted in exactly one country, and in every case that country is the one Wikidata names as the origin of a same-titled foreign work.

The obvious test does not work

The first rule we tried was “if it never charts in Korea, it is not Korean.” It is wrong, and the size of the error is the point: 342 of our 902 Korean titles — 37.9% — never chart in Korea at all. A drama that airs on KBS or tvN is on Korean television, not on Netflix Korea, so it can travel to sixty countries without ever appearing at home. Applying that rule alone would have thrown out 342 genuine titles to catch thirteen foreign ones.

So the test requires three things at once, and all three must hold:

1. The title never charted in Korea.
2. Wikidata attributes no Korean work at all to that name.
3. The market where it charted most is one Wikidata names as the origin of a work with that name.

Condition 2 is what makes it safe. A name that a Korean work also carries is never removed, however foreign the chart pattern looks.

Two we could not settle

Both met conditions 1 and 3. Both failed condition 2, so both stayed in the panel. We do not know which work charted, and we would rather carry two uncertain rows than delete two of our own.

How large the error was

The 13 titles held 84 chart places out of 37,750, which is 0.22%. They were spread across 38 of the 93 markets, and the largest effect on anything we publish was that Malta left the top five of the markets that only take wide titles and Croatia entered it.

The share is small. That is not the reason to report it, and it is not a reason to have left it alone: a method that says “we match on title text” has to say what happens when title text is not unique, and until today ours did not.

What this cannot do

It needs a foreign work that Wikidata already knows about. A title with no entry anywhere is invisible to this test, and so is one where the foreign work exists but its country is unrecorded.

It also says nothing about titles inside the panel that are correctly Korean but whose chart rows belong to a second work of the same name in some weeks and not others. We can only accept or reject a title whole. Nothing in a rank table separates one row from another.