Four European markets have almost no artists of their own in their own top 50.

One variable, two questions

We published a figure called home-grown share. It resolved an artist as local if their citizenship fell in the market’s wider language family — so Latin American artists counted as local in Spain, Brazilian artists in Portugal, German artists in Austria. It also used genre tags and ISRC prefixes to fill in artists whose citizenship we could not resolve.

That is a defensible measure of something. It is not a measure of a market’s own artists, and we were presenting it as one, in a table that ranked eleven markets against each other.

So we split it, and froze both definitions in writing before regenerating anything:

What the comparable metric says

Poland, Portugal and Sweden are absent because we could resolve citizenship for only 64.8%, 63.2% and 57.3% of their chart weight, below the 70% minimum the frozen definition sets. A number we cannot stand behind is worth less than an acknowledged gap.

The Netherlands at 0.0% and Austria at 1.0% are not errors. Dutch and Austrian listeners are not listening to Dutch and Austrian artists, in any quantity, in their own charts.

The gap between the two metrics is where the old figure was doing its work. Austria reads 1.0% domestic and 10.5% cultural-language: the difference is German acts. Spain reads 33.9% and 55.4%: the difference is largely Latin America. Neither gap is an error either, but they are not the same fact and should never have shared a column.

What we got wrong, and how

Before the split we noticed Ireland reading 0.0% on the old rule for a single quarter. We decided a measure returning zero must be broken, widened Ireland’s definition to include the United Kingdom, and cited the zero as the justification for doing so.

That is selection on the outcome. We disliked a result and changed the construct until we liked it better.

The split then showed the widening had been unnecessary as well as improper. Ireland at 0.1% is not an outlier: the Netherlands is lower, Austria is within a point, Belgium is within four. We had been about to explain away a figure that turned out to be the pattern.

Three Irish acts appeared in the Irish Top 50 across the whole year — Dermot Kennedy, CMAT and Niall Horan — and one of them accounts for nearly all of the 0.1%.

The rule we now work to

An implausible result is a reason to investigate the measurement, not permission to change the definition.

Both definitions are frozen in the repository with a version number. Changing either requires a versioned change applied to the entire reconstructed history, recorded in a changelog, and never in response to a single observation. Ireland’s cultural-language set stays flagged as having been chosen after the fact, so the audit trail outlives the embarrassment.

Method and sources

A read, not a forecast. Domestic share describes who makes the music a market listens to. It does not describe how that market feels, and five pre-registered tests of whether our music signal predicts demand have failed.