What is populated, and what is not

SPEC §7 requires song title, ISWC, copyright registration, PRO work id, songwriters, IPI, writer shares, publisher, administrator, territory and dates. None of these fields exists in any of the nine supplied files. Engine 3 (Music Asset Intelligence) therefore has no source for compositions. Populating it needs a PRO or Copyright Office connector (SPEC §11), which is Phase 2 work.

SPEC §10 requires rights holder, asset, rights type, ownership percentage, controlled share, administrator, territory, dates, source and verification status. The dataset documents none of these. The 178 artist relationships that DO exist are classified "artist_association_only" because a directory listing is what was observed. SPEC §10: ownership of an artist's discography may not be inferred from a publishing relationship, so 0 is the correct count of verified rights interests here.

SPEC §4 requires date, buyer, seller, artist, catalog, price, rights acquired, ownership percentage, structure, financing, territory and announcement URL per transaction. The news archive contains announcement text that mentions acquisitions in prose, but extracting a transaction table from it would be an LLM inference, not a documented record; doing so without evidence per field is what SPEC §26.11 forbids. Candidate posts are discoverable via news search; none has been promoted to a transaction.

A valuation needs royalty income, a disclosed multiple or a comparable transaction. None is present. The schema refuses a valuation row with no basis and no source, so the module shows no figure rather than a modelled one.

SPEC §22 task 6 asks for every publicly documented publishing catalog. The supplied files name no catalog. Playlist names such as "<Artist> Discography" are Spotify playlist titles, not catalog identifiers, and are not promoted to catalogs.

SPEC §16 decade and performance analytics depend on chart or release dates; neither is present. Requires a trade-press or chart connector (SPEC §11).

Not present in the supplied dataset. Would come from a corporate-structure source.

SPEC §4 requires revenue, EBITDA, AUM, acquisition spending, catalog valuations and royalty income, each marked reported or estimated. None is present. Primary Wave is privately held; figures would need filings or trade-press reporting, neither captured here.

Depends on composition data, which is absent. Writer shares in particular may not be estimated: the schema refuses a share with no source document.

Requires a U.S. Copyright Office or PRO connector (SPEC §11). None was used in this phase; the spec forbids bypassing access controls, so these remain Phase 2 work.

SPEC §4 requires executive management and board of directors. The supplied dataset covers the artist directory and the news archive only; no leadership page was captured.

Some news posts mention sync licensing in prose. No post gives production title, licensee, date, territory and fee as structured fields, so no placement row can be written from evidence.

All 802 posts carry at least one artist link, and their ids are exactly the distinct post ids in primarywave-artist-news.csv. How many posts exist on primarywave.com in total was not established, and this dataset cannot answer it. Posts mentioning no artist are absent.

tracks_truncated_at_100 is true for those embeds in the supplied dataset, so the 6545 track rows held are a FLOOR, not a total. Any figure derived from them must read "at least". The real playlists are longer.

playlist_owner is a third party (Spotify, an individual, or blank) on those rows. An embed on an artist page is not a curation claim, and no embed is a rights claim of any kind. Not every embed is even a playlist: entity_type includes album and track.

A playlist is not a release edition (SPEC §26.6). Playlist contents are held in playlist_tracks and are not promoted to release_tracks.

SPEC §9 requires release title, type, date, catalog number, UPC, label and territory. The playlist embeds carry tracks, not releases; 2 of 108 embeds are albums but expose no release metadata beyond a name. Engine 3 has no release source.

The dataset carries no stream counts, monthly listeners, followers or video views. Track and playlist counts are inventory of what an embed listed, and are never presented as performance.

SPEC §5 states the 900+ figure must be independently verified and not assumed. It was verified on 2026-10-22: https://primarywave.com/artists/ paginates at 20 per page over 9 pages (20x8 + 18 = 178), page/10 returns 404, and the union of distinct /artists/<slug>/ URLs across all nine pages matches this dataset's 178 slugs exactly, with zero differences in either direction. The directory can grow, so re-verify before publishing the figure again. Method and date must be displayed beside the number (SPEC §21: no completeness claim without an established universe).

data/source/PROVENANCE.md states 13,675 news posts. That is the LINE count of primarywave-news-posts.csv (13676 lines), which is inflated because body_text contains newlines inside quoted CSV fields. Parsed as records the file holds 802 posts, and primarywave-news-posts.json independently holds 802 objects with an identical set of post ids. Measured 2026-10-22 by this ingest. The figure 13,675 must not be published.

SPEC §12 forbids automatically merging low-confidence matches, so a credit stays unresolved until the resolver accepts it. Run `pnpm resolve`.