Methodology

Every number here involves two countries. A Source is a country whose chart is being read. An Origin is a country an artist is from. Only a country with a chart can be a Source. Any country can be an Origin, including one with no chart of its own: a Puerto Rican artist in Spain’s chart is credited to Puerto Rico, although Puerto Rico has no chart.

Most countries are both. Sweden is a Source when its chart is read, and an Origin when a Swedish act appears in someone else’s chart. The two are never interchangeable. “Sweden’s number” means nothing until you say which one: what Sweden’s chart is made of, or where Swedish artists are heard. Mixing them up is the easiest way to misread the map.

A Share is the fraction of one Source’s chart that belongs to one Origin on one day. Every share is relative to a single chart, and the shares of one chart add up to exactly 1, unknown origin included. Three rules decide it.

• Flat weight within a chart. Each of a chart’s N entries is worth 1/N, whatever its position. An entry at #180 in Sweden has still crossed a border, which is the thing being measured. There is no invented curve for how much more a #1 is worth, and the rule works the same for charts of different lengths.
• One country, one vote across charts. Every chart carries the same weight. Nothing is weighted by population or revenue: this measures presence across borders, not money, and weighting by the size of a country would bring back part of the home advantage that leaving out a country’s own chart exists to remove.
• Collaborations split equally among credited artists. An entry credited to KAROL G, rusowsky & Judeline gives a third of its weight to each of the three: one third to Colombia, two thirds to Spain. Every credited artist weighs the same, whatever order the entry names them in, featured acts included. That is generous but consistent. Crediting only the first-named act would erase the route by which artists from outside the West most often first reach Western charts: as a feature.

Shares are allocated in whole units of one part in a million million rather than by floating-point division, so the shares of a chart add up to exactly 1 and a chart that lost an entry can be detected.

Some credited artists can’t be given an Origin by attribution (see below): the catalogue doesn’t know the name, or several acts have it and nothing separates them. Their part of each entry goes to Unknown.

Unknown stays in the denominator. If 40% of a chart has an unknown origin, the map says 40% is unknown, and the remaining 60% is divided among the Origins that were identified. Their shares are not scaled up to fill the whole chart.

Why not drop Unknown and divide each chart among the artists that were identified? Because the artists who can’t be identified are not a random sample. Western acts are well documented in the catalogue attribution relies on, and acts from the non-Western long tail are not. Dropping Unknown would inflate exactly the Origins that are easy to identify, the Western ones, by exactly as much as attribution fails on the ones that are not.

The badge. The map shows a badge such as “17.0% unknown origin”. It is the share of chart weight with an unknown origin: each chart’s unknown share, averaged across the day’s charts with every chart weighing the same, then averaged over the days shown. It is the size of the gap in the picture. A Diet’s panel gives the same figure for that one chart.

Unknown is never drawn as a country. It is not a place, and colouring it would invent one.

Export score. An Origin’s share of every Source’s chart except its own, averaged: how much of the average foreign chart belongs to that Origin on one day. It is given in percent of a chart. An export score of 18% means that, on average, the equivalent of 18 entries in every 100 on a chart outside that Origin belong to its artists. Being an average rather than a sum, it does not grow when the archive gains charts. The default map colours every country by it, including countries without a chart. The colour scale is logarithmic, so the largest Origins don’t turn every smaller one the same pale colour, and a measured zero has its own lightest swatch.

Openness. The share of a Source’s own chart that does not come from its own Origin: 1 minus its domestic share. Unknown is part of openness, because it is not the Source’s own Origin, so a chart with a large unknown share can look more open than it is. Check the unknown share beside it.

Diet. Every share of one Source’s chart, taken together: what that chart is made of, Origin by Origin, its own Origin and Unknown included. Click a Source on the map to see its Diet. The Source itself is outlined with a neutral fill rather than coloured, so a large domestic share doesn’t dominate the scale; its domestic share, openness and unknown share are given as numbers in the panel. Only a Source has a Diet. A country with no chart has none, and the site says so rather than drawing an empty map.

Reach. One Origin’s share in each Source’s chart, taken together: where that Origin’s artists are heard. Every Origin has a Reach, including one with no chart. An export score is a Reach collapsed to one number: averaged, with the Origin’s own chart left out. On a Reach map, countries without a chart are grey, because there is no chart there for the Origin to appear in. Grey means only “no chart”, never zero.

An export score is an average over charts, and the charts it is averaged over are not the same for every country. The archive holds around 157 charts a day.

• A country with a chart has its own left out, so its export score is averaged over the other 156 or so. The one left out is the chart it would do best in, usually by a wide margin.
• A country without a chart, such as Puerto Rico, is averaged over all 157, because it has no chart of its own for the rule to leave out.

Because it is an average, how many charts there are doesn’t tilt the comparison: no country is ahead just for having one more chart to add to its score. Which charts there are still does. The rule that stops home listening from drowning the signal costs every Source its best chart, which a country without a chart never gives up. Export scores are not exactly comparable between the two.

This is stated rather than corrected. The measure is uncorrected elsewhere too: nothing weighted by population or revenue, no curve for chart position, flat weight within a chart. And this asymmetry is far smaller than the error from attribution by name, below. Correcting the smaller bias while leaving the larger one in place would make the number harder to explain without making it more honest.

The charts carry no artist identifiers. Each entry is plain text, such as “KAROL G, rusowsky & Judeline - BbY WOW”, so working out where an artist is from is done by name, not by ID.

Names are looked up in MusicBrainz, an open, editor-curated music catalogue, which gives the country or area an act is associated with: the scene it belongs to, not where its members were born. That is why Puerto Rico keeps its own code and Bad Bunny is credited to Puerto Rico rather than the United States. The whole artist field is looked up first, and split into credited artists only when it is not known whole, so Tyler, The Creator and Earth, Wind & Fire each stay one act. Where several acts have the same name, the one MusicBrainz lists under the most spellings is taken; if the leading ones are level and disagree, none is, and the name goes to Unknown. Hand-written overrides fix names the catalogue gets wrong or doesn’t have, and always win over it. Because attribution happens after the charts are collected, a fix applies to every day that artist ever charted, once the archive is derived again.

This will get some artists wrong, and the errors are not evenly distributed. Spot checks while designing the project identified 30 of 30 artists on the US chart and 10 of 10 on Nigeria’s, but only 6 of 12 on Vietnam’s. The errors concentrate in the non-Western long tail: names written in scripts and spellings the catalogue doesn’t hold, and names several acts have. That is precisely where the project’s credibility lives. An artist that can’t be identified becomes Unknown, which the badge shows. An artist credited to the wrong Origin is harder to see, and is fixed only when someone notices it.

How far a country’s music travels doesn’t shift from one day to the next, so a single day’s number is mostly noise: a release week or one viral entry moves it. By default, each date on the slider shows the 7-day average ending on that date, the date itself included. The toggle shows the single day, so you can check what one day says.

• The average is taken over the days the archive holds among those seven. If some are absent, it is taken over fewer, and the map says how many, such as “5 of 7 days”.
• A Source’s Diet and openness are averaged over the days that Source had a chart.
• An Origin with no entries in a chart on a day is a measured zero for that day, not skipped.
• Days are not weighted by how many charts they hold.
• The colour scales are fixed across the whole archive, so a colour that shifts as you move the slider is a different number, not a legend that was redrawn.

The charts are still collected every day, because nobody else keeps a daily per-country history of them: the pages they are read from are overwritten, so a day not collected is gone for good.

This measures cultural flow through the Western streaming ecosystem, not all of the world’s listening. The charts are Apple Music’s, and in much of the world Apple Music is not how most people listen. Only music is measured.

China and Russia have Apple Music charts, and their Diets are shown with a visible note that Apple Music is a minor platform there and the chart reflects Apple Music listeners only. China’s chart makes the point: around a tenth of it is Chinese, and around two fifths Taiwanese.

Iran and North Korea have no chart, so they have no Diet and are grey on a Reach map. They still get an export score whenever their artists chart elsewhere.

The charts are Apple Music’s daily per-country charts, up to 200 entries each, read once a day from kworb, which publishes them for around 157 countries. Artist origins come from MusicBrainz and the hand-written overrides described above.

Only aggregates are published. For each day the site publishes the shares, export scores, openness and unknown share, with the list of Sources and which version of the attribution the day was derived with. The chart text as collected, and the tables used to attribute it, stay private: that text is Apple’s data, passed on by kworb, and not ours to publish again. The build refuses to publish it.