What a chart position does not tell you about how much was watched

Netflix gives hours to the world and none to the countries

0 of 493,600. Not a sample, not a lag — Netflix simply does not publish how many hours a country watched anything. Every country figure on this site therefore rests on a position, and the question is how much a position carries.

How much a rank narrows the range

How this is counted. Spread is the 90th percentile divided by the 10th — how many times bigger a high week is than a low one. "Knowing nothing" is the spread across every row of a chart; "knowing the rank" is the median of the spreads inside each of the ten rank positions.

Films and series, English and non-English, are four charts of different sizes. Mixing them would let a difference between charts be read as a difference between ranks.

🔴 The row that matters to us is the highlighted one. Korean series appear on the non-English television chart, and that is where a rank narrows the range least — 45.8%, against 63.1% on the English film chart. Two weeks at the same rank on that chart are still 3.65× apart in hours.

What a rank does carry

A position is not nothing. Across every chart, a typical week at rank 1 is six to eight times the hours of a typical week at rank 10, and the ordering never inverts. Rank tells you where a title sits against the other nine. It does not tell you the size of the audience, and on the chart Korean titles live on it tells you that least well of the four.

Rank by rank, on the chart Korean titles sit on

Read the last column down. Even at rank 1 — the position with the least room above it — a high week is several times a low week. The number in the third column is a median, not a value: any single week can sit well away from it.

What one more column buys — and the trap in measuring it

If a rank alone leaves this much unsaid, the next question is what to put beside it. We tested two candidates: Netflix's own weeks in top 10, and how many countries' charts the title was on that week — a column Netflix does not publish anywhere and we build from the 493,600 country rows.

The trap first. Splitting rows into more cells narrows the range on its own, because a smaller sample puts the 90th and 10th percentiles closer together. Each real split is therefore run again against a meaningless label with the same number of buckets, and only the difference between the two is counted as information. Every figure in the table below has that sham figure subtracted from it.

🔴 On the chart Korean titles sit on, Netflix's own column is worth nothing — -0.2 points, which is to say less than a meaningless label. The column we build ourselves is worth 4.2 points, taking the chart from 45.8% to 51.3%.

Had we picked without measuring, we would have picked weeks in top 10 — it arrives in the file already and costs nothing to pass on. It is the one that does not work.