Supplier Source Profile — Supplier B

A characterization of one synthetic supplier's data as it would arrive at the platform. Profile is synthetic; structural archetype only — no real catalogue file underlies it.

Status

Authored: 2026-05-17. Last updated: 2026-05-18 with v0 decisions aligning to the layered canonical + transformation model: imprint = brand (canonical brand field), bibliographic fields handled as Print-Music-category facets per the per-category facet model, and lifecycle normalised to binary active / inactive matching Supplier A's treatment.

Stage: v0 scoping. Synthetic. Modelled on the structural archetype of a global print music publisher with an Australian distribution arm that also redistributes a stable of third-party publisher brands to AU music retailers. No real supplier file informs the contents of this profile; it captures the kind of supplier this archetype represents and the kinds of mess such a supplier produces.

This is the "medium distributor" slot in the v0 supplier set described in scope.md. The v0 scope memo originally proposed a pro-audio-shaped medium supplier (Sennheiser-like); a print music archetype was chosen instead because it covers an otherwise-uncovered v0-priority facet category (Print Music) and applies ingest-format pressure (Excel-with-human-formatting) the rich-data distributor in supplier-a-profile.md does not touch.

Identity

Synthetic ID: Supplier B.
Role: AU distribution arm of a global print music publisher. Carries the parent publisher's own catalogue and acts as exclusive AU distributor for a stable of third-party publisher brands (educational publishers, classical houses, choral specialists, method-book brands, etc.).
Catalogue size: approximately 80,000–150,000 titles across all brands, an order of magnitude larger than Supplier A. Single-brand catalogues range from a few thousand to tens of thousands of titles.

Product coverage

Sheet music, songbooks, method books, instrumental tutors, ensemble parts, choral octavos, vocal scores, exam syllabi, play-along editions with accompanying audio, and a small adjacent line of music-themed merchandise (manuscript pads, batons, etc.) depending on brand.
Brand mix: multi-brand, where "brand" in this archetype means publisher imprint. The parent publisher's own imprint is the largest single brand; a long tail of distributed publisher imprints carries the rest. Brand conventions differ — large brands carry full bibliographic detail; smaller brands carry only title + composer + format. Brand handling at the canonical layer is identical to Supplier A's multi-brand portfolio: each title's brand goes into the canonical brand field as a single string.
v0-priority category coverage: Print Music (exclusively — no other v0 supplier covers it). No overlap with Supplier A's category footprint.
Overlap with other v0 suppliers: none on shared SKUs (AU exclusive distribution holds for print music as it does for instruments). This profile exercises canonical-model breadth across a different product type, not deduplication.

Source format

Excel workbooks, one workbook per brand, delivered by email on a monthly-ish cadence with ad-hoc out-of-cycle pushes for new releases. The workbooks are human-presentable — formatted for a buyer at a retailer to skim, not for machine ingestion. Specifically:

Each workbook is a single flat sheet but carries header rows, logos or decorative banner rows, and merged cells in the top region above the actual data table.
Number formats are display-oriented (currency symbols inline with price, leading apostrophes on catalogue numbers to preserve formatting, mixed text/number in identifier columns).
Column headers vary slightly between brand workbooks — large brands use the parent publisher's standard header set; smaller brands retain their upstream-publisher conventions, so the same logical field arrives under different header strings across workbooks.
A handful of brands emit one workbook per format band (e.g. one workbook for choral octavos, a separate workbook for instrumental ensemble parts) rather than one combined workbook.

Refresh cadence: monthly bulk update emails, with new-release pushes interleaved (release-calendar driven, not FX-driven as in Supplier A).

No supplier-side push API at v0. Ingestion path is workbook upload through the per-supplier mapping config described in scope.md. Because this supplier emits many workbooks with header drift between them, the per-supplier mapping config must support multiple file variants per supplier — one variant per brand workbook — not a single supplier-wide schema. This is a mapping-editor UX consideration, not a canonical-model concern.

Field inventory

Field set is bibliographic rather than spec-based, but the canonical handling is identical to Supplier A's pattern: primary fields populate the canonical primary fields, and category-specific fields populate the curated facets for the Print Music category.

Reliably present (across brand workbooks):

Publisher catalogue number — the supplier-SKU (column header varies — "Cat. No.", "Pub. No.", "Item Code", "Product Code").
Title — the canonical product name.
Brand — the publisher imprint, sometimes a column, sometimes implicit from which workbook the row arrived in. Lands in the canonical brand field, single string, exactly as Supplier A's brand handling.
Format — book, spiral-bound, score, score + parts, octavo, digital download, audio bundle, etc. Vocabulary varies between brands. Lands as a Print Music category facet.
RRP — with currency symbol inline as a display artefact.

Intermittently present:

Composer — present on most classical, educational, and instrumental titles; sometimes carried as "Composer / Arranger" or "Composer / Artist" depending on brand. Print Music facet.
Arranger / editor — carried by classical, educational, and choral brands; usually absent on popular-music songbooks. Print Music facet.
Instrumentation / voicing — carried where meaningful (choral SATB / TTBB / SSA, ensemble scoring, solo instrument indication). Print Music facet.
Grade / level — carried by educational and method-book brands; absent elsewhere. Print Music facet.
Series — carried where the brand organises its catalogue into named series (method-book lines, "Easy Piano" collections, exam syllabus tiers). Print Music facet.
ISMN — present for newer titles from major brands; absent on older catalogue and on most smaller-brand titles. Where present, populates the canonical GTIN/barcode primary field.
ISBN — present for book-format items from publishers that registered them; sparse elsewhere. Where present, populates the canonical GTIN/barcode primary field.
Cover art — image URL or filename present for newer titles; absent for backlist items. Where present, may be a thumbnail-grade asset, not a high-resolution image. Lands in the canonical images primary field as URL pass-through (consistent with the v0 image policy).
Pages / duration — sometimes carried; can populate canonical descriptions or Print Music facets depending on curation.
Short description / contents summary — sparse; usually only for hero titles and method-book lines. Canonical descriptions field.

Notably absent:

MPN. Print music distribution does not use manufacturer part numbers. Identity for this supplier must use the supplier-SKU-fallback path named in scope.md.
Wholesale cost. Same structural absence as Supplier A — cost is per-retailer trade-discount terms applied to RRP, with schools-channel and educator-discount overlays adding further complication. v0 carries cost as a single canonical field with null for this supplier (or supplier-mapping-derived defaults), consistent with the cost-handling decision in scope.md.
Dimensions / weight. Not meaningful for sheet music products.
Marketing highlights. Not a print music convention.

Identifiers

Primary identifier: publisher catalogue number, treated as the supplier-SKU. No MPN at all for this supplier.
Supplier-SKU uniqueness within the supplier. The same catalogue number value can collide across brands — 12345 from one brand is not the same product as 12345 from another. The per-supplier ingest mapping composes a within-supplier-unique supplier-SKU (e.g., by prefixing with the brand identifier) so that the canonical supplier-SKU is unique within the supplier without requiring any special treatment at the canonical-model layer.
ISMN / ISBN intermittent. Where present, they populate the canonical GTIN/barcode field and are useful for cross-supplier matching of those specific titles. Coverage is too sparse for ISMN/ISBN to serve as primary identity across the catalogue.
Identity resolution path. Canonical-layer identity for this supplier uses the supplier-SKU-fallback path — the same path scope.md names for any supplier without reliable MPN. This is the non-trivial real-world case for that fallback.

Mess profile

This is the load-bearing section. Each item is a specific mess the canonical model or the ingestion mapping must absorb.

Excel workbooks are human-formatted. Header banner rows, merged cells above the data table, decorative imagery rows, and inline formatting (currency symbols in price cells, leading apostrophes on catalogue numbers, mixed-type identifier columns) all show up. Naive pandas.read_excel-style ingest will not find the data table without a defined header-row offset per workbook, and inline formatting will pollute parsed values.
Slightly different headers per brand workbook. "Cat. No." vs "Item Code" vs "Product Code" all denote the supplier-SKU; "Composer" vs "Composer / Arranger" vs "Composer / Artist" all carry the same logical field. The per-brand mapping variant absorbs this at ingest time. No canonical-model implication — the canonical layer never sees the raw header strings.
One supplier, many file variants. A workbook per brand means the per-supplier mapping config carries N file variants — one per brand workbook — each with its own header offset, column mapping, and inline-format scrubbing rules. The mapping editor UX must accommodate adding, editing, and removing file variants under a single supplier without forcing each variant to be registered as its own supplier.
Supplier-SKU collision across brands within the supplier. Catalogue numbers are namespaced by brand at the source. The per-supplier ingest mapping composes a unique supplier-SKU within the supplier (e.g., brand-prefixed). Same mapping mechanism that handles any namespacing decision; no canonical-layer impact.
No variant grouping in the file. Every format, arrangement, grade, and voicing of the same underlying work arrives as its own row, with no field linking them. The same work (e.g., a popular song) may appear as a solo piano arrangement, an easy piano arrangement, a piano-vocal arrangement, a guitar tab arrangement, a choral SATB arrangement, and a play-along edition with backing audio — six rows, no link. Different stress shape from Supplier A: A's variant axes are present but disagreeing; B's are absent from the file entirely. Variant grouping in the canonical layer (where retailers want it) must be inferred from title + composer + arrangement fingerprinting, or curated, or explicitly deferred post-v0.

What this synthetic supplier stress-tests, and the v0 decisions it forces or exercises

This profile stresses four concerns. Some force new v0 decisions; others exercise decisions already made for Supplier A.

Ingest format diversity (open v0 decision). Excel-with-human-formatting is a different ingest pipeline from Supplier A's CSV-via-portal. Either v0 supports Excel as a first-class ingest format (with per-workbook header offset, sheet selection, inline-format scrubbing as part of the mapping config), or v0 scope-limits to suppliers that emit structured CSV exports and treats Excel ingest as post-v0. Practically, the print music archetype cannot be ingested into v0 without Excel support; choosing this archetype as the medium supplier slot commits v0 to first-class Excel ingest.
Identity resolution via supplier-SKU-fallback (exercises existing decision). The supplier-SKU-fallback identity path named in scope.md is exercised here in its non-trivial form: no MPN at all, and supplier-SKU values that collide across brands within the supplier until the ingest mapping composes them into within-supplier-unique form. Confirms supplier-SKU-fallback is a real path the canonical layer must support cleanly, not a degenerate case.
Multi-file mapping config under one supplier (mapping-editor UX concern). A single supplier emits N workbooks (one per brand) with per-file header drift. The mapping editor must support multiple file variants per supplier and per-variant configuration. Not a canonical-model concern — the canonical layer sees only normalised post-ingest data — but a v0 mapping-editor UX requirement.
Variant grouping by inference, not by parsing (open v0 decision). Supplier A's variant axes are present-but-disagreeing; Supplier B's are absent. Whether v0 attempts inference (title + composer + arrangement fingerprint), defers to curation, or treats each row as atomic and accepts that retailers receive un-grouped variants for print music titles is an open v0 decision with direct retailer-experience consequences. Default-atomic with optional curation is the path of least implementation cost; whether retailers find that acceptable for print music specifically is a real product question.

Catalogue scale (10× Supplier A's size) is not a stress-test in its own right — it sharpens the same retailer-side filter/browse concern already addressed by the retailer UI and filter-driven exports in scope.md. Worth noting; not a new open question.

The canonical primary-fields set and the per-category facet model both absorb Supplier B's shape cleanly. Bibliographic fields (composer, arranger, instrumentation, voicing, grade, series, format) are Print Music category facets — analogous to Supplier A's spec columns (finish, body wood, pickup configuration, etc.) being Electric Guitars / Pedals facets. Per-category facet flexibility is the core design intent of the layered canonical + transformation model; Supplier B exercises that intent on the curated facet set for the Print Music category.

Update behaviour

Monthly bulk workbook refreshes per brand, delivered by email. Cadence is more predictable than Supplier A's marketing-burst pattern; volume per refresh is higher because catalogue size is larger.
New releases pushed out-of-cycle as smaller addendum workbooks or revised full workbooks, depending on brand. Release calendars cluster around back-to-school cycles and major publication seasons.
Lifecycle expressed via a status column; OOP rows often persist. Some brands emit a status column; others remove out-of-print rows from the workbook entirely; others keep the row but blank the price. The canonical layer normalises all of these to the binary active / inactive field defined in scope.md. Under the v0 per-supplier (here per-file-variant) missing-product handling setting captured in v0-scope, brand workbooks with a status column run in ignore-missing-rows mode (lifecycle derived from the source column via mapping rules), and brand workbooks that signal OOP via row deletion run in snapshot-diff mode (rows missing from a new monthly workbook → inactive). The "reprint pending" and "available while existing stock lasts" cases are not third states — they emerge from the combination of inactive status and supplier-stock-level signals, identical to Supplier A's "EOL but orderable" treatment.
Prices change less frequently than Supplier A. Print music RRPs are typically revised annually rather than quarterly. FX-driven mid-cycle adjustments are rare.
Brand additions and removals. The supplier occasionally takes on a new publisher brand (new workbook starts arriving) or drops one (workbook stops arriving). This is a structural change the ingest pipeline must accommodate without code changes — adding a brand is a per-supplier mapping-config change (a new file variant), not a deploy.

Open questions

v0 Excel ingest decision. Does v0 commit to Excel-with-human-formatting as a supported first-class ingest format? Selecting the print music archetype for the medium slot implicitly commits v0 to Excel support, but the decision should be made explicit at the v0-scope level rather than absorbed implicitly through this profile. Decision memo candidate.
Variant-grouping policy for atomic-row catalogues. Inference, curation, atomic-rows-accepted, or post-v0 deferral. Has direct user-visible consequences in retailer-facing output. Decision memo candidate.
Source-side status-column string values per brand. The canonical model is binary active / inactive (decided 2026-05-17, applied here). The remaining unknown is what discrete string values each brand workbook's status column actually carries — and for brands that signal OOP by row deletion or blank price rather than a status column, what the ingest mapping's heuristic should be. Same shape as Supplier A's source-side status open question; cheap to resolve when real catalogue files in this archetype are in hand.