15 Compliant Pain Points — Shared Across Google & Meta

Pain Point → Channel Mapping

7 Headline Modes

Modes 1–4: Meta person-matching (Andromeda). Mode 5: channel-agnostic editorial. Modes 6–7: Google keyword-matching (Search RSA + PMax text assets).

Variable-Driven Creative Model

Creative IS the targeting. Andromeda reads the creative and finds the right viewer. Self-identification at second zero drives performance.

Locked Variables

Volume Math

Template Architecture — Master Creative Blueprint (15 Slots)

Note: Happy Days remapped to ToF #3 candidate. 3 existing Canva imports are MoF comparison layouts (#6). ToF is the biggest gap.

Psychology & Pipeline by Tier

9 Creative Formats

Intent-to-Format Mapping

Production Mix Target

Power Vocabulary

Per-SKU adjective ladder. Tier 1 = pick first. Tier 2 = variation. Banned = validate_copy() error for the bound SKU.

Structural Patterns

Five converting copy structures. Match candidate to a pattern before writing — see _meta.structural_patterns in the matrix.

Market Tonality Lens v1.0 NEW · 2026-04-23

Country-first filter. Every other copy decision (mechanism, stage, mode, character budget) flows through the market lens. Resolve market first, then project downstream.

Funnel Register v2.0 4 NEW subsections · 2026-04-22

Persona-agnostic structural framework. Pain-state framing is the self-ID trigger for mechanism-bound SKUs — not aspirational/optimiser framing. Specific copy lines live in per-segment headline banks.

Tonal overlay — 2026 resilience register NEW · 2026-04-22

Register rules — Stage × SKU v3.1 · TWO-MODE TOF required / banned elements per cell

BOF three-layer stack NEW · 2026-04-22 identical for Pro + Elite

Format deltas NEW · 2026-04-22 format IS register (post-Andromeda + PMax era)

ASA-Safe Phrase Library

Green = safe to use. Red = risky — documented regulatory action or hard-to-substantiate claim.

Vocabulary Lexicon

v1.2 Extensions NEW

Added 19 Apr 2026 — EFSA / NHC authorised claims, verb hierarchy, descriptor synonyms, market-specific localisation, and the consolidated biohacking-cluster exception. NEW pills auto-hide after the _new_until date in the vault.

SA4 Short Claim Family — All 4 Markets

Primary (A) is used ~65–70%. Alt (D) is used ~30–35% when the 3-tier range moat is the angle (competitor conquest, category education, quiz funnel entry).

Banned  ·  Approved

Where it's enforced — 1 matrix + pointers

Refactored 19 Apr 2026 (Session 41 — Path 2 SSOT migration): config/brainzyme_product_matrix.json → _meta.lexicon is the single source of truth (multi-brand: each brand_product_matrix carries its own _meta.lexicon). Programmatic API: tools/product_matrix.py → lexicon() / class_word() / sa4_short_claim() / validate_copy() / authorised_claims() / verb_hierarchy(). All other files are thin pointers. This dashboard fetches data/product_lexicon.json (auto-synced from the matrix via tools/sync_lexicon_to_dashboard.py) — edit the matrix only.

1. Trigger

Calum pushed back on v1.0 funnel register framing. Two issues were raised:

2. Research Cascade (3 parallel Opus 4.7 agents)

3. Framework Changes — v1.0 → v2.0

4. Conclusions Reached

1. Framework, not templates. Register rules are structural patterns (what tone, what required elements, what banned elements) — they do NOT contain specific copy lines. Copy lines live in per-segment headline banks and get filled at commission time.
2. Pain-state is the self-ID trigger for Brainzyme. Pro audience recognises "flat / foggy / low drive" as themselves. Elite audience recognises "overwhelmed / frayed / holding too much". These are segment-agnostic — work across menopausal women, students, dads, devs, carers, founders.
3. Original is out of scope for ads. Mild entry-level formula suitable for 10–15yo children. Never appears in ad creative. Pro + Elite only.
4. 2026 resilience temper. Global economic pressure + geopolitical instability means optimisation framing reads as tone-deaf. All copy tempered toward resilience, showing up, daily consistency, getting through.
5. Format IS register. Post-Andromeda (Meta, Oct 2024) and Google PMax era. Same funnel stage needs different structure on Reels vs static vs PMax asset vs Demand Gen. Format deltas codify this.
6. BOF three-layer stack is universal. 365-day guarantee as hero (not footnote) is Brainzyme's single biggest competitive edge — Seed and Mind Lab Pro only offer 30 days. Stack order is locked, only the Layer 2 validator rotates by segment.

5. Files Updated (14)

6. Verification Cascade

7. Open Items (Non-blocking)

• banned-copy-gate.sh does not yet parse the hustle_culture_register rule. Enforcement currently via PIL copy validation + human review. Gate-level regex follow-up is optional.
• Phase 2 rollout from earlier in session (6 remaining files per copy-methodology-spec §4) still pending Calum's spec review.
• E-04 composite still shows "Gentle Focus. Every Day." Replacement ("Stress-Free Focus. Every Day.") pending — will pair with Phase 2.
• D1–D5 spec decisions (answered inline during session) not yet committed to matrix structure.

Canonical SoT: config/brainzyme_product_matrix.json → _meta.funnel_register. All other files are consumers or summaries. When in doubt, the matrix wins.

v3.0 — Why re-key by mechanism?

The funnel-register + vocab cells aren't really about the SKU label — they're about the mechanism the copy is written against. FOCUS Pro copy is dopamine-drive copy; FOCUS Elite is GABA-calm. If we ever add a second SKU to the same mechanism (e.g. a Pro+ variant), the copy rules stay identical. Keying by mechanism makes the primary key match what actually drives the copy, and makes the structure future-proof for multi-SKU-per-mechanism expansion.

v3.0 — What changed

v3.0 — Files touched

v3.0 — Decisions locked

1. Keep baseline as a first-class mechanism key. Original's cells aren't dropped — they're stored under baseline. Consumers that need Original behaviour still get a cell.
2. Audience lives on theme binding, not the lexicon. New primary_audience field is on _meta.product_theme_binding. The 4-class lexicon (classes 1–4) is contract-locked; audience is product metadata, not class lookup.
3. Compat via projection, not migration. Option B — storage is mechanism-keyed; callers can still ask for SKU view via funnel_register_by_sku() / power_vocabulary_by_sku().
4. Display columns stay SKU-labelled. Users think in SKUs; storage is keyed by what drives the copy. Renderer bridges the gap via product_theme_binding.class_3_mechanism.
5. No combo register. Starter Bundle is merchandising-only — no separate cell.
6. Bundle power_vocabulary refactor in same PR. Same schema change, same commit.

v3.0 SoT unchanged: config/brainzyme_product_matrix.json → _meta.funnel_register + _meta.power_vocabulary. The storage key moved from SKU to mechanism; the rules didn't change.

v3.1 — Trigger

User correction on v3.0 framing. v3.0 framework had ONE TOF mode — pain-state ("feeling flat / overwhelmed"). User flagged this as a cop-out: aspirational TOF is equally valid when the aspiration is mechanism-bound. Both modes co-exist:

• Pain-state self-ID: "Feeling flat? Can't get moving?" (dopamine) / "Overwhelmed? Frayed?" (GABA) — the v3.0 mode, kept.
• Aspirational self-ID: "Get your drive back." (dopamine) / "Calm under fire." (GABA) — new mode, additive.

v3.1 — Framework change

v3.1 — Headline bank additions

All 77 headlines PIL-measured (Poppins Bold) and all lexicon-validated. 0 banned_copy_patterns hits, 0 banned verbs, 0 class-4 cross-pollination. Tool: tools/pil_aspirational_bank_2026_04_23.py.

v3.1 — Files updated

v3.1 is additive only — all v3.0 pain-state content preserved verbatim inside pain_state_frame. Rollback path: restore brainzyme_product_matrix.backup-2026-04-23-v3.1-two-mode-tof.json and revert product_matrix.py/creative.html.

Research output

Full gap analysis banked at marketing/ads-creative/v3.1-research-gap-analysis-2026-04-23.md (~7 sections, Tables A-F, 22 recommended headline additions, source list). Headline findings below.

Top 5 genuine gaps (against 77 banked aspirational headlines)

Competitive differentiation — what research surfaced

Conclusion: mechanism-bound + resilience-overlay register is structurally different from the top-3 category competitors. No register collision. Ship Phase 2 with confidence.

Calum's Q1-Q4 decisions (locked)

Follow-up briefs 4 / 4 COMPLETE AWAITING CALUM REVIEW

4 out-of-scope items dispatched as parallel Opus 4.7 Sonnet agents (background). All 4 complete as of 2026-04-23. Each wrote its own spec file under marketing/ads-creative/. Phase 2 implementation (headline bank extension + matrix v3.1.1) gated on Calum reviewing all 4 before dispatch.

Phase 2 implementation (queued, not dispatched)

After the 4 briefs complete and Calum reviews, the Phase 2 implementation agent will extend the headline bank with:

• ~40 new headlines (longer Tier A 22-27 chars × aspirational × 4 markets × 2 mechanisms, plus longer pain-state example_headlines which are currently empty).
• 22 research-recommended additions — Table D from gap analysis, Q1-Q4 decisions applied (floor-framed / "all-day" / carrying/holding verbs).
• Combat-verb purge on existing GABA bank entries per Q3 REPLACE directive.
• PIL-verified at 964px @ Bold 88pt (4:5 Tier A new budget, confirmed in tools/pil_v5.py).
• Matrix version bump v3.1 → v3.1.1, dashboard sync via sync_lexicon_to_dashboard.py.

Gate: all 4 briefs reviewed by Calum + spec comments returned. Phase 2 won't touch matrix before briefs close — avoid re-ship cycles.

Cluster-Level Feedback

Give feedback at the intent cluster, market, or creative format level. Individual ad group feedback is in the Google Ads Copy tab. Use Export Feedback JSON to collect everything.

By Intent Cluster

By Market

By Creative Format

1. Dashboard Tabs — What Each One Is

Two separate dashboards. This file (v4.0-alt/creative.html) is the copy/strategy dashboard. The image-grid review dashboard lives at creative-review/index.html and loads creative-review/runs.json dynamically.

2. Canonical Library Folder Strategy

3. Pipeline Stages (FLUX → QC → PIL → Publish → Push)

FLUX 2 Pro (library_based_run / foreplay_templates_run / final_run)
    ↓  output/pending/
Stage 1.5 QC gate (publish_run.py — Claude-vision, fail-closed)
    ↓  only passes if QC declared + passed
PIL composite (pil_v5.py — v3b_highlight canonical)
    ↓  4 formats: 4:5 / 1:1 / 1.91:1 / blog
Publish (dashboard_publish.py / canonical_library_publish.py)
    ↓  creative-review/images/canonical-library/
git commit + push → apps.nutritionalproducts.org live

4. Guardrails & Hook-Enforced Gate

5. Storage Map — Where Everything Lives

6. How to Add or Modify Entries

7. Known Gaps / Things Not Yet in the Canonical Library

• ADHD blog advertorial format — System C pil_v5.composite_blog(). Separate roster, not yet commissioned into canonical library. Tracked under creative-review/images/blog-adhd-* tabs.
• Wide format missing for F-series + G-series entries — 13 of 50 roster IDs have no wide-v3 composite. Escape hatch: FLUX-native regen (right-third strict guard + per-scene constraints).
• v3c_soft variant — exists in pil_v5.py, used only for women-led + FR parapharmacie contexts. Not in the main canonical library run by default. Opt-in per-scene.
• US market creative — all current canonical entries are UK/DE/FR. US local-first copy not yet commissioned into the library.
• Flat-colour PIL variant — deferred (2026-04-15) until “text covers face/product” needs-work items are cleared.

8. Rollback Procedure

Sister-Hero Tracking — deployment report

Every banked ad composite pairs with its raw FLUX hero (the image before PIL text overlay). Composite carries baked-in PIL that cannot be un-baked — the raw hero is the only reversible anchor for future subject-swaps, PIL refreshes, or copy regens. This tab is the live status report for the 2026-04-19 sister-hero deployment.

Why this exists

Banking (tools/bank_canonical_pending.py --execute) uses shutil.copy2() — a pure file copy, no PIL re-render. When an approved ad carries a stale typography era or outdated trust-bar treatment, the only way to refresh it is to re-render from the raw hero. Before this deployment, that hero lived under a time-stamped sibling name in output/pending/ and would be pruned over time. Now it’s preserved alongside the composite with a registry that makes it findable by ID.

What shipped (6 surfaces)

Coverage backfill (2026-04-19)

Hero provenance — where the raw files came from

Workflow (new promotions)

Generate via Step 3.3 publish script
     ↓ writes to output/pending/<run>/
Run output tab (dashboard)
     ↓ click green ✓ Approve
Pending tab (virtual — computed from approvals − banked)
     ↓ click ★ Star to bank
Export Review → JSON envelope
     ↓
python tools/bank_canonical_pending.py <export> --execute
     ↓
1. Copies composite to images/canonical-library/<canonical-name>.jpg
2. Updates runs.json + banked_items.json
3. 🔗 SUBPROCESSES reconcile_canonical_heroes.py --execute
   → copies hero to images/canonical-library/heroes/hero-{ID}.jpg
   → refreshes config/canonical_hero_registry.json
4. git add — composite + hero ship in the same commit
5. git commit + push

Rescue script — when to re-run

The reconcile script is safe to re-run any time. Call it after:

• A batch of FLUX generations lands new hero-*.jpg files in output/pending/
• Manual edits to the Banked tab in runs.json
• Any suspicion that the registry has drifted (new composites were banked outside the script)

# Dry-run first — reports coverage, no changes
/c/Python314/python.exe "F:/Claude Root/tools/reconcile_canonical_heroes.py" --dry-run

# Execute — copies missing heroes, refreshes registry
/c/Python314/python.exe "F:/Claude Root/tools/reconcile_canonical_heroes.py" --execute

Downstream consumers

System Build — Headline Bank Modularization (2026-04-26)

This thread converted the 274-line curated headline bank from review artefact into a first-class matrix asset, gave copy-gen agents a typed reader API, locked the schema with hooks, and surfaced bank anchors inside the existing copy preflight. End-to-end discipline so future TOF copy generation can’t drift away from what we just curated.

What changed at a glance

Files shipped

Why this matters — the modularization story

Quick-use reference

Read curated anchors (Python)

from product_matrix import headlines, headlines_for_sku

# By mechanism (raw)
headlines("UK", "dopamine", "aspirational",
          gender="M", persona=None, min_quality=8, tier="A")

# By SKU (mechanism inferred)
headlines_for_sku("pro", "UK", "aspirational", min_quality=8)

Read curated anchors (CLI)

python tools/product_matrix.py --headlines dopamine aspirational \
    --market UK --min-quality 8 --gender M

Surface anchors during preflight

python tools/copy_preflight.py \
    --brand brainzyme --channel google_rsa --funnel tof --market uk \
    --tof-mode aspirational --sku pro \
    --copy 'Drive On. | Sharper than yesterday.' \
    --reference-headlines --ref-min-quality 8 --ref-gender M

Outstanding

• Step 7 — HerLifty matrix bootstrap. Deferred until HerLifty work begins.
• v4.0/agent-playbook.md — Not yet updated to reference Step 0 bank-anchor lookup. Flag for next session if a TOF copy run hits the playbook before this is wired in.
• Holistic system review — Plan agent identified 3 gaps; 2 are now closed (bank anchoring + matrix edits). The 3rd (brief stage unprotected) was deliberately left as doc-only because briefs are too creative to gate — revisit if compliance leaks reappear.

Campaigns — campaign_layers v1 (2026-04-26)

Time-bound or audience-bound creative overlays that compose on top of canonical lexicon and market-tonality without overriding them. First production layer: UK exam season Q2’26 (active 2026-04-15 → 2026-06-30, all 4 markets). Reader API + preflight integration + brief runner all live; hooks extended for cross-layer protection.

Why this exists

Before campaign_layers, time-bound overlays (exam season, Easter, BFCM) needed ad-hoc lexicon edits with no enforcement of expiry windows and zero compliance gating around campaign-specific claims (e.g. “pass your exams” is treatment-adjacent under MHRA/FSA — banned only in this context). This system makes seasonality declarative, auto-expiring, and compliance-gated.

Active & planned campaigns

Composition precedence (HARD RULES)

Brief workflow

1. Author a brief in campaign-briefs/<id>.json (schema in canonical memory doc).
2. Run python tools/campaign_brief_run.py <brief.json> — composes selectors, surfaces top-8 bank anchors, writes a starter pack with copy_lines TODO placeholders.
3. Writer/agent fills copy_lines.<channel>.<field>[*] in output/campaigns/<id>/starter_pack.json.
4. Lint with python tools/campaign_brief_run.py <brief.json> --check-existing → preflight_report.json (PASS / FAIL / WARN per line).
5. Active-window gate fires automatically — brief out of window aborts at step 2.

First production brief

File inventory