<!-- last-reviewed: 2026-06-01 · owner: Growth (Video Producer / Script Writer) -->

SOP-094 — Chunk tagging (HOOK/MAP/SCOREBOARD/RADIO/DELTA/PIT/CTA)

SOP · owner: Growth (Video Producer / Script Writer). Layer: SOP/Runbook (docs/canon.md). The canonical chunk list with timecodes/meanings lives in docs/concept.md (:119–130); the unit boundary lives in docs/units/growth/charter.md — not restated here. This is one of the Video Producer’s two required SOPs (artifact-matrix §3 ✓): the recurring procedure that maps every script section to a canonical chunk tag and produces the manifest Data & Analytics ingests. Mislabeled chunks corrupt the platform’s retrieval and scoring (the dual flywheel), so this SOP rules the two ambiguities Phase 0 found (reports/phase0-reviews/video-producer-codebase-review.md F1/F2).

Interface Contract

• Inputs: episode script ← from self (SOP-093) · trigger: draft complete
• Outputs: chunk-tagged manifest → to Data & Analytics Director · trigger: chunk manifest ready
• Outputs: chunk manifest → to Archivist · trigger: artifact approved
• Handoff trigger (in): an episode script is drafted (SOP-093) and ready to tag
• Handoff trigger (out): every section is tagged, the manifest is written, and it is handed to Data & Analytics for rubric scoring and library ingestion
• Definition of Done before handoff:
  • [ ] Every script section carries exactly one canonical tag (HOOK · MAP · SCOREBOARD · RADIO · DELTA · PIT · CTA), per the section→tag table below
  • [ ] The PIT/CTA ruling applied: Lap 1 and Lap 2 are [PIT]; the takeaways debrief is [CTA] (see §Rulings)
  • [ ] The 10→7 fold applied for concept chunks not in the 7-tag set (SETUP, LAP CLIPS, SETUP CHANGE, PIT WALL NOTES, ARTIFACT CTA — see §Rulings)
  • [ ] Manifest lists, per chunk: tag · section title + line-range · one-sentence summary · associated artifact path(s)
  • [ ] Any section that does not map cleanly is flagged [NOTE], not force-labelled
  • [ ] No rubric scores / quality grades added (tagging ≠ scoring — Data & Analytics’s lane)
  • [ ] Recurring unmapped patterns escalated to Data & Analytics for taxonomy review
  • [ ] Manifest handed to the Archivist (0 drift)

Triggers — registration status. artifact approved is registered (seed block); draft complete is registered (docs/handoff-triggers.md Wave-2 block, intra-unit author→self). chunk manifest ready (Video Producer → Data & Analytics) is proposed, reported to Operations (Wave-4 report), not added inline — a stub until registered. The Data & Analytics Director is a prior-wave identity (Wave 2) that does not declare this consumer edge, so it reconciles as pending (counterparty not yet authored) for this consumer relationship per interface-contract-spec.md §3.1.1 R3 — clean for this wave.

Purpose

Map every section of a drafted episode script to exactly one canonical chunk tag and produce the manifest Data & Analytics ingests — so the library’s retrieval and rubric scoring run on accurate, reproducible labels. Tagging is a contract with the platform, not a cosmetic exercise: a wrong tag silently corrupts what the platform retrieves and assembles into lessons.

When to use

Whenever an episode script is drafted (SOP-093, draft complete) and is ready for tagging — once per episode, before the script package is finalized.

The section → tag table (the canonical mapping)

Rulings (resolving the Phase-0 ambiguities — apply verbatim)

1. PIT vs. CTA (F1). [PIT] tags execution-pass chunks — Lap 1 and Lap 2 both carry [PIT] (parallel structure, the build/test reps). The concept’s “PIT WALL NOTES” (the 30s takeaways debrief, docs/concept.md:129) is not a separate tag: that content lives inside the [CTA] section (Takeaways + Artifact + call-to-action). So: two [PIT] chunks per episode; the takeaways debrief is [CTA].

2. The 10 → 7 fold (F2). The concept names ten canonical chunks (:119–130); the tagging taxonomy uses seven. The fold map (apply, do not re-derive):

   Concept chunk (docs/concept.md)	Folds into 7-tag
   HOOK	[HOOK]
   MAP	[MAP]
   SETUP (goal + constraints + tools)	[SCOREBOARD] (the goal/constraints framing)
   SCOREBOARD (rubric + test set)	[SCOREBOARD]
   LAP CLIPS (build/test montage)	[PIT] (the execution-pass chunk)
   RADIO CALL	[RADIO]
   SETUP CHANGE	[DELTA]
   DELTA (before/after metric)	[DELTA]
   PIT WALL NOTES (3 takeaways)	[CTA] (per ruling 1)
   ARTIFACT CTA	[CTA]

3. Edge cases (F5). A section with no clean mapping (extended sponsor read, title card, a recurring structural bit such as the “Sticky Hands Moment”, docs/concept.md:70) is flagged [NOTE: <reason>] in the manifest — never force-labelled. Recurring unmapped patterns are escalated to Data & Analytics for taxonomy review (do not invent a new tag inline).

Flow

flowchart LR
  A["Script drafted<br/>(draft complete)"] --> B["Tag each section<br/>(section→tag table)"]
  B --> C{"Section maps<br/>cleanly?"}
  C -->|no| N["Flag [NOTE];<br/>escalate if recurring"]
  C -->|yes| D["Write manifest entry<br/>(tag · range · summary · artifact)"]
  N --> D
  D --> E["→ Data &amp; Analytics<br/>(chunk manifest ready)"]
  D --> AR["→ Archivist"]
  classDef s fill:#1d4ed8,color:#fff,stroke:#1e3a8a
  classDef d fill:#0f766e,color:#fff,stroke:#134e4a
  class A,B,N,D,E,AR s
  class C d

Text fallback: a script is drafted → tag each section using the section→tag table → does it map cleanly? (no → flag [NOTE] and escalate if the pattern recurs; yes → proceed) → write the manifest entry (tag · section title + line-range · one-sentence summary · associated artifact path) → hand the manifest to Data & Analytics for scoring and ingestion, and to the Archivist for preservation.

Steps

1. Read the drafted script (SOP-093 output) and the concept chunk list and rulings above.
2. Tag each section per the section→tag table; apply the PIT/CTA ruling and the 10→7 fold (do not re-derive — they are settled here).
3. Handle edge cases — flag [NOTE: <reason>] for any non-mapping section; never force a label. Note recurring patterns for escalation.
4. Write the manifest at docs/content/chunks/<slug>.md: per chunk — tag · section title + line-range reference · one-sentence content summary · associated artifact path(s).
5. Verify — exactly one tag per section (or a [NOTE]), no rubric scores, no quality grades. Tagging is the contract; scoring is not yours.
6. Hand off — the manifest to Data & Analytics (chunk manifest ready) for rubric scoring and library ingestion, and to the Archivist (artifact approved).
7. Escalate recurring unmapped patterns to the Data & Analytics Director for taxonomy review.

Roles / approvers

Run by the Video Producer. Data & Analytics receives the manifest and applies the scoring rubric (their lane — never duplicated here). The Archivist preserves the manifest. The Growth Director rules on any proposed change to the 8-step format that would alter the tag mapping.

Guardrails

• Tagging ≠ scoring — apply taxonomy labels only; never add rubric scores, numeric assessments, or quality grades to the manifest.
• No forced labels — a non-mapping section is [NOTE], not a stretched tag; rubric integrity depends on accurate tags.
• Apply the rulings verbatim — the PIT/CTA ruling and the 10→7 fold are settled here; do not re-litigate per episode (capture, do not re-derive — KM discipline).
• No new tags inline — the 7-tag taxonomy is fixed; recurring unmapped patterns go to Data & Analytics for taxonomy review, never an ad-hoc tag.
• Never touch the Principal’s personal information — PA-only.
• Do not restate the concept chunk list — link to docs/concept.md.

Job aid: JA-094-chunk-tagging-cheat-sheet.md. See also: SOP-093 (script-from-brief) · docs/units/growth/charter.md · docs/concept.md · docs/handoff-triggers.md.