161 of 163 Origin Flags on This Desk’s Spotify Panel Have Confirmed the Classifier Right Since 12 September. It Was Wrong Twice

Since 12 September, this desk’s own anomaly check has flagged an artist’s registered country for a human decision 163 times across 19 collection dates of its Spotify African panel. Every one of those 163 flags got a documented check against a source, not a second look at the same registry that raised the flag. The flagged classification was confirmed correct 161 times. It was wrong twice.

19 collection dates carried at least one flag. Volume ranges from 3 to 14 and does not trend consistently up or down; the two changes sit inside the range, not at an extreme.

Show the numbers

Collection date	Flags raised
2026-09-17	14
2026-09-19	14
2026-09-20	14
2026-09-13	13
2026-09-12	12
2026-10-05	10
2026-09-21	9
2026-09-28	9
2026-10-06	8
2026-10-08	8
2026-09-22	7
2026-09-24	7
2026-09-25	7
2026-09-27	7
2026-09-26	6
2026-10-04	6
2026-10-03	5
2026-10-02	4
2026-09-29	3

Source: This desk's own panel-classification-decisions datasets, one per collection date on which the afw-data anomalies command flagged at least one artist, retrieved 8 Oct 2026 · download the data

Sponsored

161 confirmed, 2 changed, and the 2 are not where the volume is highest#

The classifier this desk’s roster runs on resolves most artists by a direct link between their MusicBrainz entry and the exact Spotify profile charting that day. Where no such link exists, it falls back to a name search, and a name search can return a different act sharing the same name, a diaspora artist whose registered country is not where they chart, or an artist two aggregators disagree on. Every one of those fallback cases gets logged as a flag and resolved by a person reading a cited source, not by re-running the same search. The full log, one row per collection date, is below.

The two changes are both named, because neither traces to first-party statement data: on 21 September, Seyi Vibez was flagged despite being a documented Nigerian artist charting only on Nigeria’s own chart, because a manual origin override had not actually been applied to his profile; applying it cleared the flag. On 28 September, Nabi’s registered country changed from Guinea to Italy, and the artist dropped out of the African panel entirely. Nabi had charted as a Guinea-credited, Italy-only export on every one of nine collection dates between 17 and 27 September before the correction landed on 28 September. Both corrections now stand in the roster this desk’s panel currently resolves against, so any dataset built from today forward carries them; the dated snapshot datasets this desk registered before each correction landed are left as the record of what was known on the day they were built, not rewritten after the fact.

What would have to be true for this to be wrong#

The headline rests on the confirmed_correct column, and it is not one date’s clean run carrying the average. Removing 17 September, the single highest-flag date, still leaves 147 confirmed-correct flags standing; removing the five highest-flag dates, which together supplied 67 of the 161 confirmations, still leaves 94. Neither of the two changes falls inside those top five dates, so the 161-to-2 split does not depend on which dates happened to get checked first.

Three limits sit underneath this count. First, it only covers dates this desk actually ran the anomalies command and found something to flag; dates between 13 August and 8 October with no registered decisions dataset are dates nothing was flagged, or the check was not run that day, and they are excluded rather than counted as zero. Second, “confirmed correct” means the registered country stood after a documented check, which is a check against a source actually read, such as a biography, a chart profile or a Wikidata entry; it is not a guarantee against an error neither the classifier nor the editor doing the check happened to catch. Third, this count stops at country-level origin. It says nothing about whether the roster’s exact Spotify-to-MusicBrainz link is correct for the 94 per cent or so of the panel that never gets flagged at all, because an artist with a correct country can still be the wrong artist entirely if two acts share a profile-adjacent name; that is a different failure mode this count was not built to catch.

What this means for reading this desk’s own numbers#

Two things follow. First, a published figure that rests on this desk’s African panel has, since 12 September, been sitting behind a human check on every flagged name, not an unexamined classifier output; the 161-to-2 record is the evidence for that claim, not an assertion of it. Second, the two misses that did get through both ran in the same direction: both were artists whose registered country understated or misstated how they actually chart, one by an origin override that had not yet been applied to a documented Nigerian artist, the other by keeping an artist on the African panel for nine readings before the classification underneath it was corrected. Readers who cited this desk’s export-column figures that counted Nabi as an African artist on any date between 17 and 27 September were citing a classification this desk has since corrected; the artist’s registered country as of 28 September is Italy, not Guinea.