The content-type family layer — the reconciliation

STATUS: OPEN — proposals 1–9 close on this page (agenda item 1). The walk itself resumes on the review pack (P-03…P-10 in its mechanism section; P-01 on its own page) — the recommendation page and the walk console are records; this page is the full reasoning behind it. The individual content types are ratified and do not move. The question on this page is only the family layer above them: what the shelves over those types are, and whether the family is a stored fact or a way of browsing. Everything ratified is shown as context; the numbered proposals at the end are what want a mark. How to mark: for each numbered proposal, answer in the walk vocabulary — CONFIRMED (the recommendation as written) · AMENDED (your wording wins) · OPEN→ALIGN (carry it) · THREAD (opens a new question). Dictation is banked verbatim before anything is processed. Prefer to see it as a system? The companion schema view renders everything on this page as a working classification table: the ontology map, one row per axis with its storage and status, the pipeline flow the ruling drives (pre-classification → routing → full classification), the four carves as structural deltas, the Codex example as actual field rows, and the build list a ruling unlocks. Prefer to see what you'd be choosing? The decision gallery shows each option as its eventual vocabulary page — the ratified types (64 after the 16 August rulings) grouped that way with live counts, your three use cases walked through, and what is literally written on each piece.

1 · What happened

At the final walk (13 August 2026) you flagged the ten family names on the ratified master list: "this proposed families list is very off… we've had a better list… am I going crazy?" The source check agreed with you on every point. The ten family names came from the machine-written recommendation section inside your own planning document and were pasted forward; you ratified the types name by name, but the family names above them were never walked. And your own ratified July model already carries a different top carve — the five-pillar output axis — that these pages never folded in. You are not going crazy; the record drifted, and this page reconciles it in one view so you can rule on the whole picture, which is what you asked for: "show me the full re-cut first."

2 · Where each family name came from

Three kinds of words are mixed in the current list, and they deserve different weight:

• Yours — carves you have written or dictated yourself, more than once, across two years of documents.
• Machine-proposed, adopted by you — wording an assistant proposed that you then re-pasted or ratified. Adopted is real, but it is worth knowing which words began as yours.
• Machine-only, never walked — wording that appeared in a recommendation and was carried forward without you ever ruling on it. This is the stratum the walk caught.

Current family	Provenance	The evidence, dated
1 · Brand & campaign storytelling	Machine-only	Never in your words. At the walk: "the word storytelling is weird here… brand identity is one thing, and then it's got visual and verbal… and then you've got campaigns" (13 Aug 2026). Your own carve keeps brand identity and campaigns apart.
2 · Product & feature communication	Machine-only	At the walk: "communication sticks out" (13 Aug 2026).
3 · Social proof	Yours (the name), machine (the family)	"Social Proof" is a category in your own hand-built content-types spreadsheet (archive, late 2025), and you picked the rename from "Proof & advocacy" on 10 Aug 2026. But at the walk you placed its members differently: "a testimonial or customer story is a content type under a product campaign, a feature campaign, or brand campaign" (13 Aug 2026).
4 · Education & support	Yours (the idea), machine (the wording)	"Educational/Informational" is a category in the same hand-built spreadsheet.
5 · Corporate communications & editorial	Machine-proposed, touched by you	You reordered the name on 10 Aug 2026; the family itself came from the machine list. At the walk: "corporate communication editorial, like that's a content category that is different than campaigns" — a live distinction you draw, not a settled family.
6 · Documentation & technical	Machine-only	Never in your carves.
7 · Advertising & promotion	Machine-only	At the walk: "it's weird to have advertising and promotion but not marketing… a commercial would live inside of a brand campaign, a product campaign, a feature campaign."
8 · Social-native & community	Machine-only, and empty	No individual types were ever named under it, so nothing in it is ratified. It is the only family with no member list.
9 · Visual & identity assets	Machine-only	You flagged it on 10 Aug 2026 as sitting right against the brand-and-identity grouping; the naming question was parked, and this reconciliation absorbs it.
10 · Event, retail & interactive experiences	Machine-only	You flagged that retail/interactive spans physical and digital; parked, absorbed here. In your 13 Aug dictation, events and experiences are one of your top worlds — a much bigger idea than a type family.

And the carve that IS yours, in your own typed words (9 July 2026): "in my head there are sort of the highest level of creative services: Brand Identity & Systems, Campaigns, Content, Digital Products & Experiences, Physical Products & Experiences — and generally most work can funnel down that hierarchy." That carve was ratified into the July classification model as the Output/Deliverable axis the very next day — and never folded into these pages. Note what you were carving when you wrote it: the highest level of creative services — what kind of work an initiative is — not a label on an individual piece. That distinction runs through everything below.

A short history of the phrase "content type" in our own records — the same two words have meant three different things across eras, which is why the lists kept fighting:

Era	What "content type" meant	Where
2025 (the touchpoint schema)	The placement's content ("Touchpoint: Radio Ad · Content Type: Entertainment")	archive, touchpoint schema, Nov 2025
Early 2026 (the big braindump and the first ChatGPT taxonomy)	"The asset format" — a 32-value list that mixed real piece kinds (hero film, product demo) with placements (OOH, paid ad, app store screenshots)	reference, early taxonomy conversations
July 2026 (ratified)	The recognizable kind of piece — judged with campaign, channel, product, role, and craft hidden	the ratified definition on the master list

One honesty note on the ratified definition itself: the sentence "Content Type identifies the recognizable kind of piece…" that appears in your 10 Aug brain dump was originally written by the assistant in the 10 July walk response, and you re-pasted it. You have adopted it — repeatedly and consistently — so it stands; but on this page it is counted as adopted, not as originated, so the provenance stays honest.

3 · Containment versus facet, in your own example

Your Codex example (from your own hand, in the planning document): a Product Campaign contains Product Launch Content, which includes a hero launch film, and a blog post on the website with a hero image showing the Codex Plugins UI and the tagline "You Can Just Build Things."

Every fact in that story survives — but it is two different structures, and the July model rules they must never be conflated:

• Containment is a relationship between objects: the campaign contains the film and the article; the article contains the hero image; the image contains a UI region. You point at containers; they have names and dates.
• Facets are descriptions each object carries at its own grain: the campaign has a campaign type (feature launch); the film has a content type (launch film) and a role (hero); the image has a content type (key visual) and a craft (graphic design); the UI region is a moment.

So the sentence composes from separate fields — Feature Launch · Hero Launch Film · Codex Plugins — instead of one nested tree "Product Campaign → Product Launch Content → Feature Content → Blog Post". The ruled consequence, stated once: content category is not above content type. The above-ness you feel is the containment chain (the campaign above its assets), and that chain is already modeled by the campaign grouping. Whatever the family layer becomes, it must not re-encode the container as a label on the piece.

4 · The candidate carves

Three candidate shapes for the family layer, plus one delta that needs your eye.

Option A — your five worlds as the family layer. Brand Identity & Systems · Campaigns · Content · Digital Products & Experiences · Physical Products & Experiences, stored as each piece's family. The intrinsic per-piece reading: which world does this piece itself serve, with the campaign name hidden. Campaign membership stays a container fact and would have to agree with the stored family (a checkable rule).

Option B — the two-level ladder from your July correction pass. The five worlds stay a lens (derived from container membership, product links, and identity-artifact flags — never stored on the piece), and the stored family layer is the seven deliverable families from the same correction: Film/Video · Static Design · Editorial/Written · Audio · Digital Experience · Physical/Spatial · Toolkit/System. Beneath them, the ladder you screenshotted: deliverable type (the ratified content types) → asset role (hero, teaser, cutdown) → variant (30s, 9:16, French).

Option C — the current ten families, kept as they stand (the superseded draft grouping).

The delta that needs your eye: your 13 August dictation of the worlds is not identical to your ratified July five. You said: "brand and identity, campaigns and content, events and experience, digital products and experiences, physical products and experiences." The July ratified five keeps Campaigns and Content separate and has no events world — events live as a grouping (event/season) alongside campaigns. Two real questions hide in that delta: whether Campaigns and Content are one world or two, and whether events-and-experiences is a top world or a container kind. This page treats the July five as the ratified baseline and flags the delta for your call (proposal 4 below).

5 · The full re-cut — every ratified type, placed under every carve

Every individual type from the ratified master list, one row each, placed under all three carves. A ⚠ means the placement is genuinely contestable — the honest finding, not a formatting accident. "Follows the piece" in the Option B column means the type is deliberately media-flexible (the ratified list says each type declares which media it applies to), so its deliverable family would change piece by piece — a case study can be a page, a document, or a film.

Legend for Option A: BI&S Brand Identity & Systems · CAMP Campaigns · CONT Content · DP&E Digital Products & Experiences · PP&E Physical Products & Experiences.

#	Ratified type	Today's family (C)	Five-world read (A)	Deliverable family (B)
1	brand film	Brand & campaign storytelling	CAMP ⚠ container-dependent	Film/Video
2	campaign film	Brand & campaign storytelling	CAMP	Film/Video
3	manifesto / anthem film	Brand & campaign storytelling	CAMP ⚠ voices identity	Film/Video
4	launch film	Brand & campaign storytelling	CAMP	Film/Video
5	teaser	Brand & campaign storytelling	CAMP	Film/Video
6	trailer	Brand & campaign storytelling	CAMP ⚠ product trailers	Film/Video
7	sizzle reel	Brand & campaign storytelling	CAMP ⚠ company reels are always-on	Film/Video
8	brand story / origin story	Brand & campaign storytelling	CONT ⚠ campaign-dependent	follows the piece
9	product film	Product & feature communication	CAMP ⚠ evergreen product films	Film/Video
10	feature overview	Product & feature communication	CONT ⚠	follows the piece
11	product demo	Product & feature communication	CONT ⚠ container-dependent	follows the piece
12	product walkthrough	Product & feature communication	CONT ⚠	follows the piece
13	explainer	Product & feature communication	CONT ⚠	follows the piece
14	product comparison	Product & feature communication	CONT	follows the piece
15	interactive demo	Product & feature communication	DP&E ⚠ marketing-content reading	Digital Experience
16	product configurator	Product & feature communication	DP&E ⚠	Digital Experience
17	app store video	Product & feature communication	CONT ⚠	Film/Video
18	testimonial	Social proof	CONT ⚠ container-dependent	follows the piece
19	customer story	Social proof	CONT ⚠ container-dependent	follows the piece
20	case study	Social proof	CONT	follows the piece
21	user story	Social proof	CONT ⚠	follows the piece
22	creator story	Social proof	CONT ⚠ creator campaigns	follows the piece
23	partner story	Social proof	CONT ⚠	follows the piece
24	expert endorsement	Social proof	CONT ⚠	follows the piece
25	review	Social proof	CONT ⚠ often earned, not owned	follows the piece
26	tutorial	Education & support	CONT	follows the piece
27	how-to	Education & support	CONT	follows the piece
28	guide	Education & support	CONT	Editorial/Written ⚠
29	lesson / course	Education & support	CONT	follows the piece
30	webinar	Education & support	CONT	Film/Video ⚠
31	FAQ	Education & support	CONT	Editorial/Written
32	support article	Education & support	CONT	Editorial/Written
33	training	Education & support	CONT	follows the piece
34	announcement article	Corporate communications & editorial	CONT ⚠ your Codex campaign contains one	Editorial/Written
35	news article	Corporate communications & editorial	CONT	Editorial/Written
36	feature article	Corporate communications & editorial	CONT	Editorial/Written
37	interview / Q&A	Corporate communications & editorial	CONT	follows the piece
38	thought leadership	Corporate communications & editorial	CONT	Editorial/Written
39	opinion	Corporate communications & editorial	CONT	Editorial/Written
40	press release	Corporate communications & editorial	CONT	Editorial/Written
41	newsletter	Corporate communications & editorial	CONT	Editorial/Written
42	research report	Corporate communications & editorial	CONT	Editorial/Written
43	technical documentation	Documentation & technical	CONT ⚠ part of the product experience	Editorial/Written
44	API documentation	Documentation & technical	CONT ⚠	Editorial/Written
45	developer guide	Documentation & technical	CONT	Editorial/Written
46	release notes	Documentation & technical	CONT	Editorial/Written
47	white paper	Documentation & technical	CONT	Editorial/Written
48	specification	Documentation & technical	CONT	Editorial/Written
49	technical reference	Documentation & technical	CONT	Editorial/Written
50	commercial / spot	Advertising & promotion	CAMP	Film/Video
51	promotional ad	Advertising & promotion	CAMP	follows the piece
53	sponsored content	Advertising & promotion	CAMP ⚠ paid-ness is a distribution fact	follows the piece
54	offer / promotion	Advertising & promotion	CAMP	follows the piece
57	key visual	Visual & identity assets	CAMP ⚠ evergreen product KVs	Static Design
58	logo animation	Visual & identity assets	BI&S	Film/Video ⚠ or Toolkit/System
59	product shot	Visual & identity assets	CONT ⚠	Static Design
60	template	Visual & identity assets	BI&S	Toolkit/System
61	brand guidelines	Visual & identity assets	BI&S	Toolkit/System
62	icon system	Visual & identity assets	BI&S	Toolkit/System
63	event film	Event, retail & interactive experiences	CAMP ⚠ events world in your Aug carve	Film/Video
64	keynote	Event, retail & interactive experiences	PP&E ⚠ its recording is content	Film/Video ⚠
65	booth / installation piece	Event, retail & interactive experiences	PP&E	Physical/Spatial
66	retail display piece	Event, retail & interactive experiences	PP&E	Physical/Spatial
67	interactive experience	Event, retail & interactive experiences	DP&E ⚠ spans physical too	Digital Experience ⚠

(Rows 52 and 56 — performance ad and public-service ad — were RULED OFF the list at the final walk, 16 August 2026: paid-vs-organic is a distribution fact, and public-service ad is retired. Row 55 — recruitment ad — was DROPPED in the next round, same date: "recruitment ad just drop"; nothing in the library carries any recruitment typing, and the employer-brand meaning lives with the employer-branding craft and the audience axes. The re-cut now places 64 types; the original row numbers are kept so earlier dictation stays traceable.)

(Family 8, social-native & community, has no ratified member types — its two illustrative examples, a creator collab and a meme format, were never ratified individually and are not counted. Its fate is proposal 3.)

6 · What the re-cut shows, measured

• Under Option A, 29 of the 64 placements carry a ⚠ — the world is genuinely contestable, and in almost every case for the same reason: the world follows the initiative the piece belongs to (its campaign or program — a container fact), not the piece itself. Your own walk sentence is the cleanest statement of it: "a testimonial or customer story is a content type under a product campaign, a feature campaign, or brand campaign." When 45% of a classification's answers depend on a fact stored elsewhere, it is a lens over that fact, not an independent axis.
• Performance ad and public-service ad were ruled off the list on 16 August; recruitment ad — the one type that had no home at all under A (employer-brand content, a world the five don't name) — was dropped at the second round the same day.
• Under Option B, 27 of the 64 types are media-flexible or contested — 22 of them literally "follows the piece" — so their deliverable family cannot be fixed at the type level without either forking types per medium (testimonial-video vs testimonial-article, the combinatorial growth the July model forbids) or deriving the family from the piece's media format, which makes it redundant with the media-format axis that already exists. The Audio family gets no ratified type at all. Where Option B is strong is exactly where the evidence for it came from: the brief archive's thirteen deliverable categories are demand-side groupings — what clients ask for — and they work beautifully there.
• Under Option C, the ten families largely restate the content-role axis — the job a piece does. Social proof ≈ prove. Education & support ≈ educate. Corporate communications ≈ announce/inform. Product & feature communication ≈ explain/demonstrate. The July model already rules content role as its own multi-valued field; a family layer that mirrors it stores the same fact twice under a different name — the same duplication defect as A, one axis over.

7 · The deliverable ladder, mapped to ruled ground

Your screenshot's ladder — output domain → deliverable family → deliverable type → asset role → variant — is not new structure; most of it is already ruled, and the July correction pass says plainly that this ladder "is different from editorial genre, channel, media format, and campaign type":

Ladder level	Ruled home
Output domain (the five worlds)	Navigation lens over the object graph (July model) — never a stored piece label; this page's proposal 1 decides whether that ruling stands or is amended
Deliverable family	The open question of this page (Option B)
Deliverable type	The ratified content types — Launch Film, Key Visual, Landing Page, Logo Animation… the same words, already ruled
Asset role (master/hero, teaser, cutdown, adaptation, template, derivative)	Ruled: the piece's job in its campaign system — stored on the campaign-membership link, not on the piece
Variant (30s, 15s, 9:16, French, TikTok, CTV)	Ruled: one creative, many appearances — renditions of the same logical creative, aspect-ratio variants included

So ruling this page's proposal 1 completes the ladder: four of its five levels already have ruled homes.

8 · Genre — parked, on your instruction

Your steer, verbatim: "i think genre falls under our creative classification which again not sure is the best word but is under more of our 'concepts' pillars… content and moments get visual and verbal classification across creative things like genre, discipline, services, executions/speciillizations i think tahts its own taxonomy track and review." The July correction pass reached the same split from the other side: narrative genre (comedic, dramatic, western, documentary-style) is tone-and-treatment vocabulary, not a second kind-of-piece axis. So: no genre axis is minted here. Genre rides the planned creative-classification review (visual/verbal: genre · discipline · services · specializations/executions), and this page only reserves the boundary: if a word describes how a piece is told rather than what kind of piece it is, it belongs to that track.

One boundary you confirmed at this review, stated once so it holds: the family is not genre. A family is a shelf that groups piece types — closer in spirit to an intent or goal grouping. Genre is creative classification: whether a piece is documentary, comedy, or cinematic in its telling. The same piece has both, on different axes, and neither word ever substitutes for the other.

9 · What the industry does

You asked for outside grounding, so a dedicated research pass (16 August 2026, eighteen systems across four worlds — the full cited write-up is in the reference record listed under Sources) asked one question everywhere: how do mature systems separate kind-of-piece from category, format, placement, genre, and campaign membership? The findings converge hard:

• Only subject matter ever earns a deep hierarchy. The advertising industry's own content taxonomy (IAB, version 3) is split into an "aboutness" tree plus flat orthogonal vectors for type/purpose, media form, environment, and source — and its version-3 breaking changes existed precisely to undo earlier mixing ("Television and Movies are now only intended to be used to mark content related to TV or Movies — not that they are TV shows or Movies themselves"). The library-science controlled-vocabulary standard states the rule formally: a hierarchy is only valid between terms of the same basic kind of concept — a discipline and a body part "cannot be related hierarchically." The ten current families fail exactly this test: some answer what job (education & support), some what form (visual & identity), some who paid (advertising & promotion).
• Campaign membership is a relationship everywhere, never a category value. The web's shared vocabulary (schema.org) models it as "is part of" — a work-to-work link; every major DAM platform models campaigns as collections an asset joins and leaves ("when a campaign wraps up, the collection can be deleted — the original files stay safe"); Meta, Google, and TikTok all store the objective on the campaign object, placement on the buy, and nothing about either on the creative.
• Paid-versus-organic is a distribution state, not a kind of piece. TikTok's Spark Ads are the proof case: an authorization code runs an unchanged organic post as a paid ad for thirty days, and it goes back to being organic when the window lapses. If "performance ad" were a content type, the same post would change type twice without a pixel changing. This independently confirms proposal 2.
• Type layers are small everywhere. Dublin Core's cross-domain type list is twelve values and has held for thirty years; the film databases run about eleven title types beside a capped genre list; the DAM vendors that commit to concrete asset-type buckets ship four to seven. Nobody ships a ten-bucket mixed layer. And the famous "seven, plus or minus two" ceiling turns out to be a debunked misreading — the honest guidance is that the count follows what distributes the types legibly, with the real tests being: does every family answer the same one question, are the families of roughly equal weight, and would a lay reader understand every name.
• The rich labels people love to browse are generated views, not stored rows. Netflix's seventy-six thousand micro-genres are sentences composed from separate clean facets ("Oscar-winning Romantic Dramas from the 1950s"); Spotify's six thousand genres are derived descriptors over listening data. That is the pattern for your five worlds: compose the world from facts the model already stores, and it can be as rich as you like without anyone maintaining it by hand.
• One borrowable marking aid: the film databases label each genre as objective (rule-decidable) or subjective (a judgment call). The re-cut table's ⚠ marks are exactly that split — the un-flagged placements are rule-decidable, the flagged ones are judgment.

10 · Routing and prompt impact, per option

The family layer is not cosmetic: the pipeline's router chooses which prompt a piece gets in two steps — first by where the piece came from, then by its content grouping — and the ratified governance rule says every classifier's choice list is generated from the registry. A live probe (16 August 2026) makes the stakes concrete: the router's first layer has sixteen active routes by source, and its second, semantic layer exists but is switched off — twelve rows, every one inactive, keyed on the superseded mixed carve (brand campaign, product campaign, brand identity, customer story, social content…). What re-keys or retires that dormant layer is exactly this ruling; the schema view walks the full flow. So whatever the family layer becomes:

• Option A (worlds stored on the piece): the router's second step would key on a value that 29 of 64 types can't answer without looking at the container — so routing would effectively depend on campaign membership, which is often unknown at ingest time (a piece is usually classified before anyone knows its campaign). Prompts asking the analyzer to pick a world would be asking it to guess a container fact from pixels. This is the expensive option to operate.
• Option B (deliverable families stored): the router's second step would approximately restate the media-format split the router already makes at its first step (the analyzer is already format-driven: image, video, text each have their own envelope). Low routing value; the generated choice lists would mostly mirror the format filter that exists.
• Option C (the ten kept): routable, but the generated choice lists inherit the role-duplication — analyzers would be offered family words ("social proof") that compete with the content-role vocabulary they are also asked to emit, the exact both-axes-offer-the-same-word drift the registry exists to prevent.
• Family as registry grouping only (the recommendation below): routing keys stay on the two facts that are knowable at ingest (source and media format) plus the type once classified; family never enters a prompt at all — it only groups the choice list visually and drives browse shelves. Cheapest to operate, nothing to backfill, nothing for the analyzer to guess.

11 · The architect's pushback, stated plainly

You asked for this directly: "if organizing content categories based on how i think makes sense in my head isnt right from a prodcut database architect cto perpsective i need that pushback."

1. Your five worlds are real — as the top of the WORK, not of the piece. You dictated them as "the highest level of creative services," and that is exactly what they are: what kind of initiative or engagement work belongs to. Campaign membership, brand-identity programs, product experiences — the model already stores every one of these as containers, groupings, or product links. Storing the world again on each piece duplicates a container fact, drifts from it silently, and encodes how teams are organized (brand team vs campaign team vs product team — your own second-pass framing) into a piece's identity. Pieces outlive re-orgs.
2. The ten families duplicate the role axis (section 6). Same defect, different axis.
3. The deliverable families duplicate the format axis for the 27 media-flexible types, and the evidence base for them is demand-side (briefs — what clients ask for), which is its own model.
4. Therefore the family layer should carry no new stored fact at all. Every candidate carve duplicates an axis that already exists — world duplicates container, role-families duplicate role, deliverable families duplicate format. The stored fact remains the type (one primary, several allowed, exactly as ratified). The family is a parent pointer in the type registry — the same pattern the moments vocabulary already uses, where shelves group moment types without any moment storing a shelf. Rename freely, regroup freely, nothing to migrate, nothing drifts.
5. Small flags for the record, not re-litigated here: "app store video" carries a touchpoint in its name (the piece is a product video; the app store is where it runs) — worth a note in the definitions queue, not a re-opening. "Sponsored content" and "review" both lean on ownership (paid-ness, earned-ness) — their definitions should say the type stands regardless of who paid.

12 · Proposals awaiting your mark

Proposal 1 — the top carve. Status, 16 August 2026: you took the MECHANISM — provisionally — in your dictated walk ("I want to take your recommendation, but a few call-outs"). The mechanism (family = registry shelf stored on no piece; the worlds = computed lenses) is treated as taken pending your final word; the SHELF NAMES are now your own carve, applied on the recommendation preview. At your second round you named the identity shelf Brand & Identity and its depth review is running; Advertising & Promotion stayed dissolved; Recruitment Ad was dropped. Your final word on that page closes this proposal. The recommendation as originally written: the five worlds become derived lenses; the family layer is registry grouping only; the stored per-piece fact remains the type. Concretely — (a) each of the five worlds is defined as a saved view composed from facts the model already stores: Campaigns = pieces whose container is a campaign; Brand Identity & Systems = pieces whose type is an identity artifact or whose container is a brand-identity program; Content = pieces with no initiative container (your "always-on world"); Digital/Physical Products & Experiences = pieces attached to product experiences and physical builds. (b) A conformance check enforces the agreement you asked for: a piece's world-lens reading and its container facts must never contradict. (c) The families over the type list are re-grouped for browsing under your worlds' names where they fit (identity artifacts under Brand identity & systems; event/retail/interactive under Experiences), and the remaining role-flavored families (social proof, education & support, editorial) stay as browse shelves — labels on the registry tree, stored on no piece. If you would rather store the world on every piece (Option A as written), the routing cost in section 10 is the honest price.

Proposal 2 — re-home "performance ad." Status: RULED, 16 August 2026 — your words: "that's like organic vs paid, not a content type, wrong part of our ontology/schema/taxonomy." Recorded on the decided page; the migration of its typed pieces is pipeline-lane work. Remove it from the type list. Whether a piece is paid is ruled onto ownership, placement, and ad-format; a "performance ad" is a promotional ad (or a commercial, or a key visual) that ran in a paid placement with a conversion objective. The ad platforms' own model agrees — the same creative runs organic and paid without changing kind. Migration: the handful of pieces typed performance-ad re-type to their real kind; paid-ness lands on the distribution facts.

Proposal 3 — the empty family. "Social-native & community" has no ratified members. Either name the types you actually want (creator collab and meme format as first-class piece kinds — they pass the ratified test: an editor would call a piece "a meme" with the channel hidden), or retire the family and let those pieces take their real types with social-native carried by channel and treatment. Recommendation: name the two types, drop the family (they shelve fine under storytelling/social-proof groupings).

Proposal 4 — the worlds delta. Your August dictation differs from your ratified July five: campaigns-and-content merged, and an events-and-experiences world added. Recommendation: keep the July five as the lens set (campaigns and content are different worlds precisely because "content" is your always-on world with no initiative container), and treat events as what the model already made them — a container kind (event/season) that any world's pieces can attach to — rather than a sixth world. If your instinct says events deserve a world, that is an amendment to ratified ground and it is yours to make.

Proposal 5 — the two parked family names. Families 9 and 10's naming question (parked on 10 August) dissolves under proposal 1: their member types regroup under Brand identity & systems and Experiences lenses, and the old names retire with the family layer. A mark on proposal 1 closes this one automatically; it is listed so nothing silently disappears.

Proposal 6 — ongoing piles like Customer Stories. The live test case grew: FOUR customer-story piles exist at the 17 August re-probe (30, 9, 8 and 4 pieces), each stored as a campaign with the type customer-story-campaign. An outside review argued piles like this deserve their own grouping kind ("content program / series"); your reaction was that "ongoing advocacy program" is not how this industry talks, and that these are still campaigns — always-on ones, tied to the brand or classifying down by product and feature. The always-on idea was in fact already seeded in the campaign-type work from the brief archive, though it never reached the live registry. Three ways to mark it: (a) always-on campaign — keep such piles as campaigns and add always-on to the campaign-type vocabulary (your instinct, and the industry's word) · (b) a new grouping kind — program/series joins campaign, brand-identity program, project, and event/season · (c) both — always-on campaign-type now, and revisit a program kind only if real piles refuse to fit. The recommendation is (a): it matches how you and the industry already talk, changes one vocabulary instead of the grouping model, and amends nothing you have ruled.

Proposal 7 — what to call a version of the same creative. First, the plain ground, because "work" landed badly as a word: a content item stays exactly what you said it is — the smallest atomic unit (an Instagram image post, a YouTube video, a hero image on a landing page, a tagline on a product page), each one fully tagged on its own, which is the product. The only thing above it is a LINK: when the 16:9 YouTube upload and the 9:16 cutdown are the same creative, a link between those items is what lets you ask "show me this film everywhere it ran." You already ruled that link exists (one creative, many appearances, aspect-ratio variants included). The outside review suggests naming the middle more precisely: a 30-second 9:16 French cut is a version of the creative; the YouTube upload is an appearance of that version; our scrape is the evidence of the appearance. The pick, and it is only a naming pick: (a) keep two levels as ruled (work → appearances) · (b) name the middle level (work → version → appearance), which helps dedup, metrics, and rights later. Recommendation: (b), as naming only — nothing about the atomic content item or its per-item tagging changes either way.

Proposal 8 — one type or several, confirmed in your own words. You flagged "secondary type" as sounding wrong, and your example is exactly the rule already in force: a screen-recording of a feature being used is a whole piece whose type is product demo; a product demonstration inside an iPhone hero commercial is a moment, never the commercial's type. The only question left is the one you ruled in August and are welcome to re-confirm or amend: may a piece carry MORE than one whole-piece type — a piece that is genuinely a testimonial AND a product demo end to end — with one marked primary? (a) CONFIRM several-with-one-primary (your August ruling; the second type must describe the whole piece, never a moment inside it) · (b) AMEND to one type only (a second-type link stays addable later without loss). Either way the moment rule stands untouched.

Proposal 9 — where digital-product content lives (your second-round question, 16 August). Your words: "should we probably put Digital Product somewhere… like where do we put that ChatGPT like ios and or web app etc — re content we scrape from it and classify it." The evidence, probed: the product reference file behind the products layer is a name registry — brands mapped to product names — with no content linkage at all; ChatGPT appears in it only as a product name. So product-ness gives a piece no kind. Two ways to mark it: (a) product-ness is a LINK, never a family — a screen captured from the ChatGPT app keeps its real type; which product it came from is the existing content-to-product link; where in the product it appeared is the channel stack's touchpoint level (an owned digital-product surface — app screen, web-app screen), which the ratified stack already anticipates; and if raw interface captures need a kind of their own, ONE new type — product interface screen — joins the Interactive & Digital Experience shelf · (b) a stored Digital Product family over such pieces — which would restate the source and touchpoint facts as a second stored value, the same duplication defect the re-cut measured on the other stored-family options. The recommendation is (a): it answers your question with machinery the model already has, and it adds at most one honest type.

Sources

Everything above traces to these records, by path and date, so you can see your own words in context:

• the final-walk dictation — docs/reference/2026-08/2026-08-13/final-walk-round1.md
• your brain dump — docs/reference/2026-08/2026-08-10/content-type-brain-dump.md
• the walk response it re-pastes — docs/reference/2026-07/2026-07-10/walk-response-juuly10-01.md
• your five-worlds origin, typed — docs/reference/2026-07/2026-07-09/firstconvoref.md (the "highest level of creative services" passage)
• the ratified July model and its correction pass — docs/specs/PRODUCT/CANONICAL-CLASSIFICATION-MODEL.md §0 · docs/specs/DATABASE/WALK-FINDINGS-2026-07-10B.md §2
• your review dictations of 26 and 21 July — docs/reference/2026-07/2026-07-26/atlas-review-responses.md · docs/reference/2026-07/2026-07-21/theratificationcnooslejuly21.md
• your hand-built content-types spreadsheet and category sheets — archive/toreview/bt-content-types-ref.csv · archive/toreview/classification-ref-visual-verbal/
• the early taxonomy conversations — docs/reference/misc-ref/joebig-braindump.md · docs/reference/2026-07/2026-07-08/chatgpt-taxonomy-good-longconvo.md
• the brief-archive deliverables taxonomy — docs/specs/PRODUCT/TAXONOMY/DELIVERABLES-AND-CAMPAIGN-TAXONOMY.md
• the BRANDTRACKERS 5.0 "Content Category / Type" page and the 3.0 Taxonomy Hub — read live via the docs connection, 16 Aug 2026 (the 6.0 schema sketch of March 2026 leaves its Content Type list blank)
• the industry research pass, every external quote cited — docs/reference/2026-08/2026-08-16/content-taxonomy-best-practices-research.md
• the ratified master list this page reconciles — the content-type master list