PMax Cascade Go-Live Dashboard v1.21 — ARCH v4.19 LIVE · cascade scaffolded all-the-way-down · 5 new brief trees + deploy worklist

Phase 5.5: T1/T3/T4/T5 NO-GO (working the blockers); T2 reinstated 2026-05-16 — Phase 5.5 QC pending

← Command Centre · v1.16 (2026-05-17) · PHASE 5.5 PANEL + PATH B SELECTED + THREE-STATE DECISIONS + A/B + QC TIER · Phase 5.5 ↓ · Go-Live Journey → · Financial Audit v2.28 · Decision Board

❓ Go-Live Buckets — Questions for Calum → Go-Live Journey → Financial Audit v2.29 Decision Board

Session 25 — Go-Live + Wave 2 task progress (updated 2026-05-17)

Currently working on (2026-05-17): T-Zero brand-refresh scoped + live-audited (8 campaigns, £690/d — see T-Zero panel below). Full cascade review done — cycle-1 is NOT go-live-ready: all 4 tests blocked, Customer Match import broken (UK+DE, ~478 failed jobs). #111 done; #112/#113/#109 still await Calum. Pending questions are consolidated in the Go-Live Buckets page.

# Task Status Note
101 Build arch-targeting-registry.json DONE Built + Codex-QC-refined + committed (dbe5b95)
102 retro-c2 — strip stimulant-absence claims + commit DONE Committed d9c8499
103 cfa dashboard — bank retro verdicts DONE BANKED state shipped
104 Audience-signals logging — investigate + recommend DONE Recommendation delivered
105 Verify retro-m2 CM segment sizes ≥1k DONE retro-m2 verified — new CM segments all ≥1k across 4 markets
106 retro-h6 — PMax specs to Maximize Conversions DONE Committed 5ffa63c (Wave 2 pass)
107 retro-h3 — log cycle-2 ≥GBP100/d PMax budget DONE Committed 5ffa63c (Wave 2 pass)
108 Action go-live decisions H-R into cascade specs DONE Committed 5ffa63c — 80 files, 3 QC rounds
109 T2 single-arm scope BLOCKED Needs Calum’s top-parent-video pick
110 T4 URL expansion — 5 FR product handles PENDING Phase 5.5 blocker — ready to start
111 T3 ENABLE-CHECKLIST spec-drift fix PENDING Phase 5.5 blocker — ready to start
112 LP image fills — welcome-back-v4 + subscribe-save-v4 PENDING Phase 5.5 blocker — ready to start
113 Resolve 36 go-live campaign rows BLOCKED Awaiting Calum’s per-row verdicts
114 cfa dashboard restructure PENDING Lower priority
115 Go-live one-button unification PENDING Optional
116 Re-validate bz_* CM lists PENDING Scheduled 2026-05-17

⚠️ Cross-cutting advisory — PMax × narrow-audience channel mismatch

Per audience-segments reference: PMax cannot hard-fence a Customer Match list (signal only — expands to similar converters). 12 BRAND-section AGs across 4 markets bind narrow retargeting audiences (cart abandoners / lapsed / subscription reminders). 7 standalone PMax campaigns in config/campaign-architecture.json share the same flaw. Codex 5.5 + Gemini 3.1 Pro parallel QC (2026-05-18) refined the channel plan — both flagged DG's ~1,000 active-matched-user serve minimum and the loss of RLSA Search retargeting PMax was capturing.

✓ STATUS 2026-05-18: architecture v4.14 → v4.15 SHIPPED — all 7 arch rows in pmax_campaigns[] mutated (channel + channel_pre_v4_15 + _decision_2026_05_18 + channel_knobs_v4_15 + RLSA parallel-layer notes; pmax_us_lapsed flipped to status:POOLED_INTO_UK). All 12 BRAND-section AGs in this dashboard's ROWS data updated (channel field + channel_updated_v4_15 flag); affected rows now highlighted green in Section A below with strike-through on the prior PMAX value and the v4.15 rationale inline. Cascade strategy + tactics may need refresh given the channel change — that's the next step.

QC-refined per-campaign plan (banked, see go-live-buckets v1.14):

AG type Markets Primary channel Secondary (parallel) Critical knobs
LAPSED CUSTOMERS — large list (UK)UK (~3,300)DGRLSA on UK brand-SearchOT off, no Lookalikes
LAPSED CUSTOMERS — small listDE (~860) · FR (~700)DisplayRLSA on per-market brand-SearchBelow DG 1k floor → Display retargeting only
LAPSED CUSTOMERS — US (standalone Display)US (~400)DISPLAY (US)Standalone US-store Display campaignv4.19 update (Calum 2026-05-18): US-pool absorption REMOVED (geo contradiction). New arch row pmax_us_lapsed_standalone: US store, USD, US fulfilment. Display has no audience-size floor for serving. RLSA layer on US brand-Search 13522300536.
CART ABANDONERSUK · DE · FR · US (~1,300 global)DisplayDG fallback if eligible (Codex)Dynamic product banners; cheapest path; 7-day underdelivery trigger
SUBSCRIPTION REMINDERAll markets (~18k purchasers 30-60d)DG—OT off
T5 SUB UPSELL (cycle-1)UK (15,361 bz_2plus_purchases_12mo_no_sub)DGRLSA on UK brand-SearchOT off, no Lookalikes, 7-day frequency checkpoint @ £143/d

NOT flagged: T-Zero brand PMax (4 markets) — broad brand audience, PMax-signal fine. T-ADHD + Section-C revival — intent-keyword Search (right channel already). Student campaigns (T1/T2/T4) — prospecting / broad audience, PMax-signal fine. Decision card banked at go-live-buckets channel_mismatch_re_route.

📊 Phase 4: Cycle-1 Cascade Tests — Path B selected (4 retained tests)

Phase 4: Cycle-1 Cascade Tests — PATH B SELECTED

UPDATE 2026-05-16: T2 reinstated as a single arm (top parent videos) — see the Phase 5.5 panel. The 2026-05-14 Path B record below stands as history; T2's structural reconciliation (A/B table, UTM, AB_TESTS metadata) is tracked in task #93. #65 remains dropped.

Locked 2026-05-14: Path B = reduced cycle-1 scope (4 retained tests instead of 6). T1 / T3 / T4 single-arm B-only (drop A baseline; v4.0 LPs only). T5 keeps A/B at £143/d budget LOCKED (Card I). T2 + #65 DROPPED_CYCLE1. Cycle-1 dispatched and pre-flight verified via new cascade-qc-gate skill (Phase 2 SHIPPED). Status: PAUSED → ENABLED transition pending Calum CEP approval on 4 retained tests once welcome-back-v4 + subscribe-and-save-v4 LP builds land.

T1 UK Student PMax Single-arm B-only · £45/d · 11 UK city LPs arch_id: pmax_uk_attention
T3 UK LAPSED PMax Single-arm B-only · £15/d · welcome-back-v4 (Option C URL-param) arch_id: pmax_uk_lapsed (parent)
T4 FR Student PMax Single-arm B-only · 3 FR city LPs (Paris/Lyon/Lille) arch_id: pmax_fr_anti_fatigue
T5 UK SUB Upsell PMax A/B preserved · £143/d (Card I) · 42-56d window LOCKED arch_id: pmax_uk_brand
T2 UK DG-Parent Reinstated 2026-05-16 · single-arm (top parent videos) · scoping task #93 arch_id: dg_uk_parent (pending)
#65 SERP UK Exam-Season Search DROPPED_CYCLE1 — PMax custom intent covers coverage
—
Brand Lifecycle (PMax) — T5 A/B · T3 single-arm
—
Student — T1/T2/T4 single-arm · #65 dropped
5
Audience-led DG/META
Mid-funnel SEM
32
Total AGs/campaigns
In-flight sub-agents (2026-05-14 21:45) — historical · click to expand
21:30 CLAUDE welcome-back-v4 LP build — Option C URL-param dynamic modules (per-cohort routing: ELITE / PRO / ORIGINAL). T3 B-side dependency.
20:30 CLAUDE subscribe-and-save-v4 LP build — T5 B-side dependency. Hero copy locked: "You already love Brainzyme. Now save 16%". Audience signal bz_2plus_purchases_12mo_no_sub (2,765 UK members).
21:45 CLAUDE Multi-market stimulant-claim sweep — revise 11 UK LPs to "no synthetic stimulants" framing (Option A locked) + FR/US/DE compliance check.

Phase 4 blockers — before ENABLED transition

CALUM Card A ASA fix on 4 evergreen product pages (in flight, parallel session tasks #50-#59).
CALUM Card G image library finalisation (in progress).
CLAUDE welcome-back-v4 LP build (in flight, see above).
CLAUDE subscribe-and-save-v4 LP build (in flight, see above).
CLAUDE 11-LP "no synthetic stimulants" template fix (sub-agent in flight; Option A locked).
CLAUDE T5 LP URLs (post-T5 build; placeholder until subscribe-and-save-v4 lands).
CEP 4 retained tests PAUSED → ENABLED (Calum CEP approves, Claude executes via Gate B cascade-qc-gate skill).
v1.8–v1.10 history (deep links + three-state + feedback-as-decision)

v1.10 (2026-05-13 22:00): typed feedback in any row's textarea ALSO counts as substantive submission — no verdict required if feedback is non-empty.

v1.8: 5 per-row deep-link buttons (A→ / B→ / Tactics / Strategy / Intel) + per-row LP-live-status chips pulled from pages_registry.

v1.7: 7 Pending Decision cards + per-row History/Feedback buttons.

v1.4: ~840KB cascade brief content under projects/briefs/cascade-2026-05-13/. 7 sub-cascades.

v1.1: v1.0's 37 AGs collapsed to 16 brand-lifecycle AGs (4 per market). Parents-of-Teens routed PMax → DG.

v1.10 three-state decisions + feedback-as-decision: Each verdict button cycles 3 states. Suggested (dashed, dimmed) = Claude's pre-pop, not counted. Selected (filled tint) = clicked once, still not counted. Confirmed (solid + check mark) = double-tap to confirm; counts as decision. NEW v1.10: typed feedback in any row's textarea ALSO counts as substantive submission — no verdict required if feedback is non-empty. Submit envelope includes both confirmed verdicts AND feedback-only rows (deduplicated by row).
📍 Pending Decisions — moved to Go-Live Buckets

📍 Pending Decisions — moved to Go-Live Buckets →

These pending decisions have moved to the Go-Live Buckets page — the questions hub (5 buckets + decision cards). The cards below are kept as a faint reference trail only; answer them in Go-Live Buckets, not here.
🎯 A/B Cycle 1 LIVE — 5 tests (T5 A/B; T1-T4 single-arm) · #65 dropped

🎯 A/B Cycle 1 LIVE — 5 tests (T5 A/B; T1-T4 single-arm) · #65 dropped

Mechanism: 6 campaign-level UTM tracking templates with ValueTrack substitution at click time (per reference/utm-tracking-templates.md v2.1). NOT per-page URL baking. Sheet rows A107:Z112 in tab 2026 Launch UTMs. {_bztest} custom param identifies each test (TEST-UK-STUDENT-01, TEST-UK-DGPARENT-02, TEST-UK-LAPSED-03, TEST-FR-STUDENT-04, TEST-UK-SUB-05).

TestCampaignA (existing)B (v4.0)Mechanismarch_id
T1UK Student PMaxDROPPED AB-only: 11 UK city LPs /pages/exam-season-{city}Single-arm PMax (Path B)pmax_uk_attention
T2 (reinstated)UK DG-ParentDROPPED AB-only: top parent videos (pick pending #109)Single-arm DG (reinstated 2026-05-16) · video pick pending #109dg_uk_parent (pending)
T3 (α reframe)UK LAPSED PMaxDROPPED AB-only: /pages/welcome-back-v4 (Option C URL-param)Single-arm PMax (Path B)pmax_uk_lapsed (parent)
T4FR Student PMaxDROPPED AB-only: 3 FR city LPs /pages/exam-season-{ville}Single-arm PMax (Path B)pmax_fr_anti_fatigue
T5 (A/B retained)UK SUB Upsell PMax/pages/subscriptions (PageFly)/pages/subscribe-and-save-v4 (in build)PMax Experiment 50/50 · 42-56d £143/d (Card I 2026-05-15)pmax_uk_brand
#65 DROPPEDSERP UK Exam-Season Students/products/focus-supplements-for-students-pro—DROPPED_CYCLE1 (PMax intent covers)—

Path B mechanics (locked 2026-05-14): T1 / T3 / T4 run v4.0 B-side only (no A/B). T5 retains A/B (only one cycle-1 test with a stable A baseline). T5 budget locked £143/d (Card I 2026-05-15) × 42-56d for conversion-volume floor. #65 dropped from cycle-1 scope; T2 reinstated single-arm 2026-05-16. Phase 2 Gate B (cascade-qc-gate skill) gates PAUSED → ENABLED. Spec details: projects/briefs/cascade-2026-05-13/.

Cascade brief index (post Path-B 2026-05-14) — click through to see full specs per campaign:
  • T1 UK Student PMax — projects/briefs/cascade-2026-05-13/uk-student/ (10 files, single-arm B-only, £45/d signal investment, 11 UK city LPs) RETAINED
  • T2 UK DG-Parent (single-arm) — projects/briefs/cascade-2026-05-13/uk-dg-parent-light/ (reinstated 2026-05-16 — single-arm top parent videos; scoping task #93) PENDING SCOPE
  • T4 FR Student combined #2+#3 — projects/briefs/cascade-2026-05-13/fr-student/ (15 files, single-arm B-only, 3 FR city LPs) RETAINED
  • DE Student light #4 — projects/briefs/cascade-2026-05-13/de-student-light/ (9 files; "Ohne Crash" fact-fix applied; signal-only for 2027 build) SIGNAL-ONLY
  • T3 UK Brand Lifecycle (LAPSED) — projects/briefs/cascade-2026-05-13/uk-brand-lifecycle/ (23 files; single-arm B-only welcome-back-v4 Option C) RETAINED
  • T5 UK SUB Upsell PMax — new test (subscribe-and-save-v4 build in flight; A/B retained 42-56d £143/d, Card I) RETAINED
  • LP template enhancement — projects/briefs/cascade-2026-05-13/lp-template-enhancement/ (6 files; ~380 lines new Liquid = 46× leverage)
  • Parent-variant LPs + 4 top-city — projects/briefs/cascade-2026-05-13/parent-variant-lps/ DROPPED_CYCLE1
  • #65 SERP UK Exam-Season Search — projects/briefs/cascade-2026-05-13/uk-student-search-65/ DROPPED_CYCLE1
Outstanding Calum decisions (cycle-1 unresolved):
  1. FR DG video hook 3: "Pas de Ritaline" vs backup "Pas de stimulants" REVIEW
  2. FR PMax H4: keep "Médecins" or swap "Pros de Santé" (broader/safer) REVIEW
  3. FR PMax budget: £18/d (subagent default) vs £30/d (Codex+Gemini learning-floor) REVIEW
  4. DE/FR/US brand-lifecycle cascade (12 more AGs) — dispatch now or wait? REVIEW

Resolved 2026-05-14: Path A/B/C re-scope → B · T5 window → 42-56d · UK tier diff → Option A (cycle-1) · welcome-back-v4 design → Option C URL-param · stimulant-absence claim → Option A "no synthetic stimulants" · FR bz_2plus_purchases_12mo_no_sub → 540d (renamed 18mo) · ARC convention Q1 → CONSOLIDATE · Tracker arch_id Q2 → ADD COLUMN · Cascade overrides Q3 → BYPASS.

🎯 Phase 5.5 Targeting + Intent QC — Wave 1

🎯 Phase 5.5 Targeting + Intent QC — Wave 1

Wave 1 verdict (2026-05-14): the 4 tests assessed in Wave 1 (T1/T3/T4/T5) returned NO-GO with documented HARD BLOCKs. T2 was reinstated 2026-05-16 — Phase 5.5 QC still pending. Cycle-1 ENABLE is paused awaiting Calum sign-off on the blockers below. Per-test breakdown drops down on click; per-test comment box pushes to inbox.nutritionalproducts.org/jobs for Claude.
T1 NO-GO T3 NO-GO T4 NO-GO T5 NO-GO T2 QC PENDING Framework: ops-targeting-qc Phase 5.5 (49-check, PMax-aware)
Wave 2 — Codex 5.5 + Gemini Pro adversarial review (2026-05-15): both reviewers independently confirm all 4 NO-GO verdicts STAND. Codex verdict REFINE; Gemini verdict AGREE. Wave 2 surfaced HIGH findings wave-1 missed — mostly Google Ads control-plane mechanics — plus 6 decisions for Calum.
⚠ Convergent HIGH findings (both reviewers) — click to expand
  • T3 multi-AG PMax is mechanically unsound — PMax does not enforce per-asset-group budgets / tROAS, and does not hard-gate audiences to AGs (NEW AG bleeds into LAPSED/CART). Needs structural rework.
  • T5 A/B is statistically underpowered — 42-56d @ £25/d yields ~35 conv/variant vs 150 needed. Not fixable by "reframe"; needs budget push, metric change, or defer — resolved: T5 budget pushed to £143/d, Card I 2026-05-15.
  • Medication-comparison intent must be pre-upload human-gated — "natural alt to study drugs" (T1) + Ritaline-adjacent FR (T4) reclassified REVIEW → HARD pending Calum sign-off.
  • PMax audience signals are not hard fences — T5 Customer Match must be set as "Targeting"; all other PMax audiences as "Signal".
  • Cross-campaign T3↔T5 SUB cannibalisation — both target subscription-adjacent warm UK audiences; needs a hard pre-enable overlap check.
  • Bidding-strategy spec mismatch (Codex) — T1/T3/T4 pair Maximize Conversions with tROAS; tROAS belongs with Maximize Conversion Value. T4 FR restricted-keyword cert (Gemini) — if the FR account lacks supplement cert, zero ads serve at all.
6 decisions for Calum (respond via the per-test comment boxes below): 1. T3 structure — separate campaigns vs negative-audience exclusions · 2. T5 — push budget £143/d vs change metric vs defer · 3. T1/T4 medication-comparison intent — remove vs sign off · 4. T4 budget £18 vs £30 · 5. T4 FR cert verification · 6. 18-24 exclusion — keep vs parent-frame vs separate arms.
Synthesis: projects/briefs/cascade-2026-05-13/phase-5-5-wave-2/wave-2-synthesis.md
⚡ T-Zero — Legacy Brand Campaign Asset Refresh

⚡ T-Zero — Legacy Brand Campaign Asset Refresh (NEW workstream — Calum 2026-05-17)

What it is: the 8 existing legacy brand campaigns — 4 brand PMax + 4 brand Search — brought into the launch for a full asset + video refresh. Brand creative is 1.5–2 years stale fleet-wide (the US PMax asset group is literally named "old asset group"). Targeting and budgets unchanged — creative refresh only.

Status: SCOPED + LIVE-AUDITED 2026-05-17 — 6 scope decisions locked; 8 refresh streams defined. Next: the creative-production cascade (8 streams) — not go-live-blocking, the brand campaigns already run.

MarketBrand PMax (ex Legacy Smart Shopping)Brand Search (SERP k/w: brainzyme)
UK17485671252 — £180/d816611992 — £30/d
DE21771203454 — £250/d14357494269 — £30/d
FR21787222699 — £100/d13542095497 — £30/d
US17882283348 — £40/d13522300536 — £30/d

Live budget: £690/day total (PMax £570 + Search £120). Scope doc: projects/briefs/t-zero-brand-pmax-refresh-2026-05-17/scope.md

🧠 T-ADHD — ADHD Campaign + DE/FR/US Expansion

🧠 T-ADHD — ADHD Campaign + DE/FR/US Expansion (NEW workstream — Calum 2026-05-17)

What it is: the existing UK ADHD Search campaign brought into the go-live, plus 3 net-new market versions. UK — "SERP UK ADHD" (11666978147, Search, £60/d, ENABLED, working well) — tracked as-is. DE / FR / US — net-new builds mirroring the UK 2-ad-group structure ("natural / non-prescription alternative" framing).

Compliance is the gating factor — rules differ per market: UK/US "natural alternative to study/smart drugs" eligible (RED relaxation); DE ADHS / FR TDAH permitted (territorial override). Never an ADHD-treatment claim. Each new market runs the campaign-strategy gate + per-market copy preflight (ADHD = human-gated AMBER/RED).

Status: SCOPED + UK-AUDITED + QC'd 2026-05-17 — Codex + Gemini plan-QC'd (both REFINE, fixed). 4 open questions for Calum in the scope doc. Scope: projects/briefs/t-adhd-campaign-expansion-2026-05-17/scope.md

Section A — Brand Lifecycle (16 PMax AGs across 4 markets). The core daily-spend fleet. Each market gets ONE PMax campaign with 4 AGs mapped to lifecycle stages: NEW → CART → LAPSED → SUB. Audiences pre-exist in 12 of 16 (UK/FR/US lapsed wired; DE lapsed audience exists, AG rebuilt). Cart Abandoner + Subscription Reminder are net-new audience builds.

Section A — Brand Lifecycle Fleet + T-Zero (16 PMax AGs incl. 4 T-Zero brand PMax · + 4 T-Zero brand Search)

Mkt Role AG name (+ ID) Audience signal Channel £/d ROAS tgt Signal Intent Compl Action Fin A/B test Recommended Your decision Feedback Hx/FB & Links

Section B — Student Season (T1 UK + T2 UK DG + T4 FR, all single-arm · #65 Search dropped)

v1.2 calendar correction (Calum 13 May): 2026 verdict is per-market, not blanket-DEFER. Today = 13 May 2026. UK GO NOW (A-levels 12 May - 25 June + GCSEs 13 May - 19 June = 5+ weeks live window starting today). FR GO (Bac philosophie 16 June + Bac général through 24 June = 5-6 weeks lead time, ideal for proper build). DE LIGHT TEST (Bayern Abitur peaked this week; ~2-week late-Länder tail — small £8/d test to catch tail + validate for 2027). US DEFER TO AUTUMN 2026 (AP done 16 May; autumn SAT Oct+Dec + college midterms Oct-Nov is the real next opportunity — build now for September launch, not 2027). DG-Parent campaigns: GO for FR (parents in decision window), TEST for UK (parents mid-cycle), DEFER for DE/US (no awareness runway 2026).
Mkt Role Campaign Audience Channel £/d ROAS tgt Signal Intent Compl Action Fin A/B test Recommended Your decision Feedback Hx/FB & Links

Section C — ADHD Campaigns (UK + DE/FR/US expansion)

Section C is now the ADHD campaigns (per Calum 2026-05-17). The UK ADHD Search campaign (live, re-segmented 2026-05-16) plus the new proposed DE / FR / US versions (T-ADHD expansion). The audience-led DG/SEM rows previously here are parked on hold — retained in the data, not rendered. ADHD compliance is per-market: UK/US natural-alternative framing eligible; DE ADHS / FR TDAH permitted per the territorial override; never an ADHD-treatment claim. T-ADHD scope: projects/briefs/t-adhd-campaign-expansion-2026-05-17/scope.md.
Row Mkt Lane Audience signal Channel £/d ROAS tgt Signal Intent Compl Action Fin A/B test Recommended Your decision Feedback Hx/FB & Links
Not in this dashboard (and why): The 15 ROWS_PMAX_TOPIC rows in Financial Audit v2 (UK GABA, UK Dopamine, DE GABA, etc.) are topic-specific PMax campaigns, not part of the brand-lifecycle fleet. They have kw-intent layers (people search “GABA supplement”), so PMax is the right channel for them. Decide on those in Financial Audit v2.20 Wave 3+4. The 6 ROWS_LIFECYCLE rows in v2.20 are now superseded by Section A above — check v1.1 here for the canonical action.
confirmed: 0 / 30 · feedback: 0 · v1.10: typed feedback also submits

Spec viewer

Loading...