Proposed — the classification schema, whole

STATUS: PROPOSED (10 August 2026). This page is the entire classification system in one view — every vocabulary, at every grain, with every proposed change marked in place. It exists because reviewing the vocabularies one flat list at a time hid the structure: the same idea kept re-appearing at different levels under different names, and you could not see it without holding four pages in your head at once. Every value and count here was probed against the live database or read from the ratified pages on 10 August 2026. The moment section is the piece under active review — it carries the promotion round and the registry cleanup as one drawn-out proposal. The promotion list remains the detailed record behind it.

How to mark it: the marks legend is — RATIFIED (already canon, shown for context) · KEEP (no change proposed) · PROPOSE (a change this page asks for) · OPEN CALL (numbered OC1–OC8; your pick is needed). Strike or amend anything; an unstruck PROPOSE executes in the pipeline lane after ratification.

The words this page uses, once

Five kinds of word appear below, and most past confusion came from mixing them:

• A grain is the thing being classified: a content item, a moment, an entity, or a grouping.
• An axis is one question asked of that thing ("what kind of piece is this?", "how was it shot?").
• A level is a step inside one axis's hierarchy (a family above its types; a group above its shelves).
• A value is one answer on an axis (launch_film, establishing).
• A vocabulary is the set of values an axis or level may use — curated (a governed list) or open (new words allowed, reviewed on a schedule).
• A display name is how a value renders to a person ("Launch Film") — stored values never appear in the interface.

The rule that untangles everything: an axis answers exactly one question. When one word answers two questions — or two axes answer the same one — that is a defect, and every one found is flagged below.

The map

This map draws the four grains — the things classification attaches to — not the model's pillars. Products and concepts are not missing: products live on their own model chapter, and the classification axes themselves are what the concepts pillar governs.

Each grain carries its own axes. Below: the content item's, the moment's (in full — the piece under review), then the entity's and the grouping's.

---

Grain one — the content item

What kind of piece it is — RATIFIED

The content-type master list binds at the TYPE level: the industry-recognizable kind of piece, judged at the piece itself, several allowed with exactly one primary. The ten families it listed are shown next for comparison — the family layer is re-opened (the ⚠ below).

1. Brand & campaign storytelling · 2. Product & feature communication · 3. Social proof · 4. Education & support · 5. Corporate communications & editorial · 6. Documentation & technical · 7. Advertising & promotion · 8. Social-native & community · 9. Visual & identity assets · 10. Event, retail & interactive experiences

⚠ The family layer above was RE-OPENED at the final review (13 August 2026) — the types stand (64 after the 16 August rulings); the ten family groupings are the superseded draft, kept for comparison. The reconciliation is built and being walked: the recommendation page is the ONE review surface, with the reconciliation page behind it. The families-9-and-10 naming question is absorbed into it. See the review agenda, item 1. PROPOSE (a pending migration, made visible by this page): the live registry still stores the pre-ratification names for families 3 and 5 (proof_advocacy · editorial_corporate_comms) — the rename to the ratified words runs with the vocabulary migration.

The ratified split rule travels with it: the old mixed "category" field divides into subject scope (what it is about) · campaign type (a fact about the grouping, never a label on the piece) · content role (the job it does).

Where it ran — RATIFIED

The channel stack binds: five levels, each answering one question — Media type (eight: digital · broadcast · print · streaming & OTT · outdoor · retail · physical products & experiences · digital products & experiences) → Channel (the lane within it) → Platform (the named service, linked to its entity when tracked) → Touchpoint (where a person meets it) → Ad format & placement (the bought-media unit). The ten-value registered channel list retires into the stack; the derived six-value list in use today maps in.

OC1 — a name collision this page surfaced. The registry carries a second, different axis also called media_type — seven values about production form: live_action · animation · motion_graphics · static_image · mixed_media · audio_only · interactive. That is a real and useful axis, but it answers "how was this made?", not "what medium did it run in?" — and after the stack lands, two axes would share one name. Proposal: rename the production axis production_form. Your call on the name.

Who it's for, and why — the strategy axes

Axis	The question	Vocabulary (all curated)	Mark
Content role	What job does the piece do?	19 values — awareness · education · proof · conversion · onboarding · thought leadership · comparison · demonstration · retention …	KEEP
Marketing intent	What is the marketing move?	12 values	KEEP
Funnel stage	Where in the funnel?	6 values — awareness · consideration · conversion · onboarding · retention · expansion	KEEP
Audience segment	Who is it aimed at?	10 values — consumer · business · developer · creator · enterprise · smb · startup · education · government · healthcare	KEEP as the starter — the walk ruled audience becomes a real object with a beyond-tech vocabulary, owed by the Concepts chapter
Audience goal	What should that audience do?	5 values — learn · engage · purchase · advocate · retain	KEEP
Value-prop clarity	How clearly does the piece land its value?	7 values — high · medium · low · too technical · too vague · demo-dependent · requires trust proof	KEEP
Subject scope	What is it about — brand, product, feature, customer, partner?	the existing content level axis (brand · product line · product · feature · ecosystem · experience) is today's storage for most of this question; the ratified split absorbs it and adds the customer/partner scope	PROPOSE (part of the ratified split — content level folds in, it does not survive as a second axis)

Three thinner curated axes also live at this grain — duration category (6 values) · story structure (7) · messaging type (8) — lightly used so far; they ride along unchanged.

The craft read — what the analyzer sees in the making

These axes are emitted by the analyzer per asset and write to their own tables. The registry is their governed home; the ruled endgame (already decided, queued for the pipeline lane) is that classifier prompts are generated from the registry, and every classification carries the taxonomy version that produced it.

Axis	The question	Vocabulary	Mark
Concept	The creative idea or device	~60 curated values in nine idea-families	KEEP
Writing style	The register of the copy or script	11 curated	KEEP
Medium	What the craft sits on — live action, photography, interface capture, computer graphics	6 curated	KEEP — overlaps the production-form axis in OC1 (live_action appears on both); the migration reconciles the two so the value lives in exactly one
Technique	The execution move — match cut, parallax, kinetic type …	10 curated	KEEP
Material	The finish — chrome, matte, photoreal, textured …	7 curated	KEEP
Behavior	How things move or are treated — reveal, loop, morph, grain, lighting …	14 curated (9 motion + 5 treatment)	KEEP
Element	The design part on screen — logo, typography, color, iconography, data visualisation …	12 curated (+ 8 type styles)	KEEP
Expression	Six fixed poles — form, render, energy, register, mood, weight — each rated every time	6 poles, fixed bands	KEEP
Principle	The design principle at work	21 curated	KEEP
Product surface	Which product surface is shown — chat, dashboard, code editor, packaging …	10 curated	KEEP
Content origin	Human, AI-generated, AI-assisted — and the generating tool when identifiable	4 closed + 10 soft tool names	KEEP

What it says — the message axes

Axis	The question	Vocabulary	Mark
Claims	What the piece asserts, with claim type + how it is evidenced	65 curated claim types, harvested then promoted	KEEP
Proof points	What backs the claims	30 curated	KEEP
Risk signals	What could cut against the brand	30 curated	KEEP
Themes	The subjects and ideas it touches	open graph — deliberately unbounded	KEEP
Features	Which product capabilities it shows	open graph, tied to the product tree	KEEP
Tone · visual style · composition · pacing · color · music	The sensory reads	small curated lists each	KEEP

Who made it, and who appears in it

Crafts — the discipline tree: 95 active disciplines under 11 top-level crafts, multi-parent by ruling (one display parent). Every active discipline carries a hand-authored rulebook — inclusion rules, exclusion rules, confusable pairs, examples (one further rulebook survives on a retired discipline) — which the Concepts chapter renders into this site as a generated review page (your directive; some definitions "still a little off"). Credit roles — 65 curated roles across 11 departments, written on credits.

Entity role in content — how an entity figures in a piece: 13 curated values — primary subject · owner · mentioned · partner · customer · competitor · sponsor · talent · production credit · storyline element · demonstration tool · incidental · background. This is the mention-side read (who appears); it is a different fact from the canonical ownership chain (whose work it is — the pending backfill the conformance report measures). KEEP.

---

Grain two — the moment · THE PIECE UNDER REVIEW

A moment is one beat inside a video. It is described by one filing axis and a set of side axes, and they are two different kinds of thing: the filing axis has three levels (where the moment goes on the shelf); the side axes are independent questions you filter by. Past confusion came from reading a side axis as if it were the filing, so here is one real moment threaded through both:

A two-second closing beat where the logo assembles over a black card: filed under group Brand & Motion → shelf Logo Animations → detail logo_morph · described as subject brand marks in motion · story job sign-off · framing reveal.

One beat, every question answered exactly once. That is the target. The current registry does not fully hit it — the proposals below make it true.

The filing axis — Group → Shelf → Detail, as proposed

One axis, three levels, one question: "what kind of beat is this, and where does it file?" The group is a heading for navigation (never written on a moment — derived from the shelf). The shelf is the answer — the kind of beat. The detail is the curated craft vocabulary beneath each shelf — the promotion round's 64 values placed where they live, with the tail merges collapsed under their survivors. (The shelf says what kind of beat; the story-job side axis below says what the beat does for the story — "Product Demo" is a kind, "demo" as a story job is a purpose, and a Title Card beat can serve a demo story job.) Uses in parentheses; the promotion list carries every definition and merge line individually.

Product — the beat is about the product

• Product Reveal — hero_reveal (32) · chip_reveal (10)
• Feature Highlight — real_time_translation (8) · voice_mode (8) · display_reveal (7) · ai_integration (6) · heel_max_air_unit (5, OC2 below)
• Product Demo — ui_walkthrough (166) · screen_capture (56) · voice_interaction (41) · vfx_breakdown (28) · ui_interaction (19) · chat_interaction (17) · transformation (14) · crafting_process (11) · ai_generated_character (10) · code_generation_comparison (9) · code_editor_ui (9) · image_generation (9) · code_generation (9) · metamorphosis (8) · ai_agent_navigation (8) · ai_chatbot_interaction (8) · voice_search_response (8) · auto_framing (7) · voice_command_text_edit (7) · data_analysis (6) · ai_analysis (6) · tennis_swing (5, OC2) · race_car_in_action (5, OC2) · ai_translation (5) · component_assembly (5) · ai_product_description (5)
• Product in Use — lifestyle (165) · lifestyle_with_product (29) · product_detail (14) · hands_on_interaction (12) · typing_on_laptop (11) · tennis_challenge (10, OC2) · laptop_use (7) · active_use (7) · athlete_in_action (5) · writing_in_notebook (5) · hands_on_use (5) · person_using_phone (5)
• PROPOSE — flatten the nested shelf: product_demonstration currently sits under product_demo as a shelf beneath a shelf — the one place the shelf level nests. The ruling already says flatten it; it merges into Product Demo.

Interfaces

• UI — chat_prompt_ui (97) · progress_animation (12) · panel_ui (10) · file_browser_ui (8) · dashboard (7) · voice_app_ui (6) · task_selection_ui (5)

Typography & Titles

• Title Card — full_page_title (160) · text_overlay (86)
• End Card — logo_lockup (7)
• PROPOSE — the ruled sub-splits, drawn in: the walk ruled typography needs sub-splitting into full-screen title animations · end cards with pricing or a call to action · opening titles. Proposal: full_page_title promotes as the full-screen-titles home under Title Card; a curated pricing_cta_card detail lands under End Card (absorbing the work a stranded registry value has been failing to do — see OC5 below); opening_title is minted empty, ready for harvest.

Brand & Motion

• Logo Animations — logo_animation (52) · logo_morph (6)

Design

• Data Visualization — stat_reveal (15) · comparison_chart (13)

Editorial

• Transitions — (no promoted detail yet — 10 uses total at the shelf level)

Effects & VFX — OC3: the group exists with no shelf under it. Proposal: keep the group (the effects vocabulary is coming work, not a mistake), and note it stays invisible in browse until a shelf lands.

Shot Types — OC4, the biggest structural call on this page: a group named "Shot Types" holds two catalogue shelves — Dialogue (talking_head 933 · speaker_presentation 7 · in_scene_dialogue 5 · reaction 5, see OC7) and Establishing (establishing_shot 187) — while a whole separate Framing axis (below) answers "how was it shot?" with its own establishing, interview, reaction values. The same words at two levels is exactly the defect this page exists to remove. Proposal: dissolve the Shot Types group; Dialogue and Establishing remain as shelves whose visibility label says catalogue (kept, searchable, never offered as creative reference — the ruling's own definition); the detail value establishing_shot retires into its shelf (it adds no information the shelf does not carry); talking_head stays — it is the canonical dialogue detail. How the beat was filmed is then answered in exactly one place: the Framing axis.

OC7 — three promoted details that duplicate Framing values: reaction (5, under Dialogue) · lifestyle (165, under Product in Use) · product_detail (14, under Product in Use) each also exist as Framing values. A moment filed Product in Use and framed lifestyle says everything the lifestyle detail says. Proposal: retire all three into the Framing axis — the rows keep their shelf and their framing, and each question is answered once. If you would rather keep any of them as shelf-specific craft words, strike that line.

PROPOSE — tier becomes a visibility label (ruled, drawn in): shelves currently carry a three-way tier (shelf / catalogue / lens). The ruling replaces this with one job per list: every shelf is a shelf, and a separate visibility label says browse or catalogue. OC5: the one lens-tier value, pricing_cta, is parentless and unreachable today; under this schema its job is done by the pricing_cta_card detail under End Card plus the story-job value call to action — the lens value retires.

PROPOSE — the two shelf-name echoes retire from the detail level: product_in_use (4 uses) and end_card (2 uses) were typed into the detail level; the shelf already records what they tried to say.

OC6 — three same-idea pairs inside the promoted list (promote both, or merge now): active_use · hands_on_use — chat_interaction · ai_chatbot_interaction — data_analysis · ai_analysis.

OC2 — four values that clear the floor but name a campaign's subject, not a kind of beat: tennis_challenge · tennis_swing · race_car_in_action · heel_max_air_unit. Promote or park; the honest reading is they belong to subject/theme axes.

The three main side axes — what you filter by

Independent questions, never part of the filing. This is where the promotion page's section headings came from — they were the Subject axis, not the groups. The two lists share words on purpose-adjacent ideas (subject product vs group Product; subject typography vs group Typography & Titles) but answer different questions, and they can disagree: a beat of someone jogging with the shoe barely visible files under Product in Use while its subject is people. The group is where it files; the subject is what is literally on screen.

Axis	The question	Vocabulary (curated, complete)
Subject	What is literally on screen	product · people · environment · brand marks in motion · evidence and charts · typography
Story job	What the beat does for the story	hook · problem · reveal · demo · proof · transformation · trust · call to action · sign-off · explanation · other
Framing	How it was shot	establishing · hero · product detail · macro · lifestyle · reaction · reveal · demonstration · interview · point of view · b-roll · transition

OC8 — words shared across axes, on purpose or not: reveal is both a story job and a framing; demo (story job) sits near the Product Demo shelf; transformation is both a story job and a promoted detail. These are different questions that can honestly share a word — a reveal-framed shot usually serves a reveal story job — but if you want every axis's vocabulary to be unique words, say so and the pipeline lane renames on migration. Proposal: accept the sharing (the axis a value sits on is always explicit in the data), flag it here once, and move on.

The rest of the moment's reads

Beyond the filing and the three main side axes, a classified moment carries further descriptive reads — same grain, smaller questions:

Read	The question	Vocabulary
Human presence	Who is on screen	individual · group · crowd · none
Product visibility	How present the product is	in hand · device · interface shown · implied · absent (the ledger's agenda item 5, raised 11 August, proposes MERGING this row and the next into one axis — mark there, not here)
Product role	What the product does in the beat	focal · active use · contextual · implied · background · absent
Offering form	What kind of offering appears	16 curated values
Moment-level role & surface	The content-role and product-surface reads, at beat grain	mirror the content-item axes above
The open reads	Motion state · interest type · reuse value · before/after · visual theme · on-screen text · brand-asset flag	open or yes/no — recorded, not governed lists

Beneath the shelves, the detail level stays open — harvest-then-promote is the ruled loop: new words land freely; on a schedule, values above the floor promote with definitions and one-off spellings merge into their curated neighbor. This round is that loop's first run.

---

Grain three — the entity

Axis	The question	Vocabulary	Mark
Kind	What sort of thing	company · person · song · artist	PROPOSE — retire the fifth registered kind, music_label (zero rows use it; labels are companies); the four in use stand
Company type	What sort of company	17 curated — brand owner · sub-brand · product line · agency · creative studio · production company · production studio · holding company · media company · publisher · technology · vendor · music label · non-profit · government agency · trade association · award organization	KEEP (the walk's two-level brand-vs-vendor read composes from these)
Industry position	Sector → industry → category trees	the market trees, own tables	KEEP
Positioning — audience	Who the entity aims at	the same 10-value starter list as the content grain's audience axis — one vocabulary, used at both grains	KEEP as the starter; the ruled real-object, beyond-tech vocabulary is owed by Concepts (one owed list covers both grains)
Positioning — market	Segment and tier	consumer · prosumer · smb · enterprise / luxury · premium · mass market · value	KEEP
Positioning — occasion	The occasion it positions around	ruled first-class; no storage exists yet — honestly unbuilt	PROPOSE (build with the vocabulary migration)

The product tree beneath an entity — products are entities with their own identity (ruled), and they carry their own small axes: product type (3 curated values) · product category (20 curated) · feature type (3 curated, on the feature tree). KEEP — these grow with the vendor-side collection.

Ownership and brand architecture are relationships, not classifications — they live on the entities pillar.

Grain four — the grouping

Axis	The question	Vocabulary	Mark
Grouping kind	Which of the four ruled kinds	campaign · brand-and-identity program · project · event/season — stored today as 21 registered campaign types (product launch · feature launch · brand platform · seasonal · sponsorship · partnership · project · …), every value in use registered	KEEP — the 21 types read as sub-kinds of campaign plus project; how the four ruled kinds map onto storage is pipeline-lane design, flagged there
Membership role	What a piece is doing in its grouping	asset · hero · supporting · validation · partnership · reaction · feature	KEEP
Company role	What a company is doing in a campaign	9 curated — partner · customer · sponsor · partner brand · creative agency · media agency · production · influencer · other	KEEP (the vendor-side collection leans on this)
Event category	What kind of event or occasion	a two-level tree: 8 categories over 52 subcategories, on the events-and-observances calendar	KEEP — reconcile with the occasion axis when it is built, so "occasion" is one idea, not two
Event participation	How an entity relates to an event	14 curated roles across sponsorship · speaking · hosting · attendance · jury · award	KEEP

---

The governed surface, stated once

The registry is the single governed surface (ruled): every curated vocabulary above lives in it, classifier prompts are generated from it, and every classification records the taxonomy version that produced it. Today five vocabularies are still hardcoded in prompts and kept in sync by hand — the known split-brain defect, fix ruled and queued. Ratifying this page is therefore ratifying what the analyzers will be told to see, not just what the site displays.

Two vocabularies ride on every classification, whatever the axis: evidence — how sure the read is (directly shown · directly stated · strongly implied · weakly implied · speculative · no reference) — and who said it (analyzer · human · import · rule). The ruling binds them everywhere: every classification records who said it, and automation never overwrites a human row.

Every proposal on this page, in one list

1. Registry rename to the ratified family names for content-type families 3 and 5 (migration).
2. Subject-scope storage lands per the ratified category split, absorbing the existing content-level axis (migration).
3. OC1 — rename the production-form axis so "media type" means exactly one thing; reconcile its overlap with the craft Medium axis in the same move.
4. Moments: flatten the nested product_demonstration shelf (ruled).
5. Moments: tier → visibility label, one job per list (ruled).
6. OC4 — dissolve the Shot Types group; catalogue visibility carries Dialogue and Establishing; establishing_shot retires into its shelf; framing questions answered only by the Framing axis.
7. OC5 — retire the unreachable pricing_cta lens value; its job moves to the pricing_cta_card detail + the call to action story job.
8. Typography sub-splits drawn in (ruled): full-screen titles · pricing/CTA end cards · opening titles.
9. The promotion round (details on its page): 64 values clear the floor, 76 merges, 2 shelf-name retirements — less whatever items 6, 10, 11 and 12 below remove from the promoted set (establishing_shot retires under OC4; OC7 would retire three more; OC2 may park four).
10. OC7 — the three promoted details that duplicate Framing values (reaction · lifestyle · product_detail): retire into Framing, or keep any you strike.
11. OC6 — the three same-idea pairs: merge or keep both.
12. OC2 — the four subject-word values: promote or park.
13. OC8 — words shared across axes (reveal · demo · transformation): accept the sharing, or rename for uniqueness.
14. OC3 — keep the empty Effects & VFX group as coming work.
15. Entity: retire the unused fifth kind registration.
16. Occasion storage is built when the vocabulary migrations run (ruled first-class); reconcile with the event-category tree so "occasion" stays one idea.

Everything here executes in the pipeline lane after your marks — nothing on this page changes the database by being written down.