Artefact 15 · Step 5 / 7 · Fast Synth (No Real OASIS)
MiroFish Synth
Fast-synth persona-walkthrough for the 8 templates × 4 named personas = 32 resonance cells. This is the agent's projected verdict, not the OASIS sim verdict — MiroFish backend was down at survey time. Step 5 real-OASIS validation deferred until backend restored.
Backend status: MiroFish OASIS not responding at http://localhost:5001/api/projects at survey time (2026-05-27). Steps 5b (real-sim) and 5c (interview-mine) are deferred. Artefacts 19 + 20 are placeholders pending backend restart. Below is the fast-synth verdict from the orchestrator (Claude) — directional, not behavioural-grounded.
15.1 Resonance Matrix · 8 Templates × 4 Personas
Projected resonance score (0-10) per cell. Aggregates: copy fit (vocabulary match) + occasion fit (when persona would order) + objection neutralisation (how well the template counters the persona's top objection) + channel match (does the template's lead channel match the persona's channel mix).
Template
K-Drama Aficionado
Urban Office Worker
Halal Family Provider
Social Gatherer
Row mean
T1 · Fried-to-Order
8.4
8.2
7.6
6.8
7.8
T2 · Ambassador Co-Sign
9.1
6.4
4.2
8.7
7.1
T3 · Chimaek Daypart 🏆
9.2
7.0
3.6
9.4
7.3
T4 · Per-Piece Math (Merdeka)
6.6
8.4
8.9
6.0
7.5
T5 · XHS Bridge
8.6
8.1
3.0
6.4
6.5
T6 · Spice-Ladder Dare
8.0
5.8
2.8
9.1
6.4
T7 · Live-Event Tie-In
8.9
6.5
3.8
9.3
7.1
T8 · Reunion-Math Bucket
4.8
8.6
9.0
6.2
7.2
Column mean
7.9
7.4
5.4
7.7
7.1
15.2 Top-Resonance Cells (≥ 8.5)
Eight cells score ≥8.5. Each = a paid-media bet that should compound month-over-month. Allocate ~70% of the first 90 days' paid spend across these cells.
T3 × Social Gatherer · 9.4 — Chimaek Hour for concert / fan-meet crowds. The clearest GMV unlock of the survey.
T3 × K-Drama Aficionado · 9.2 — Late-night drama-binge daypart. Persona pain stack #2-3 directly addressed.
T5 × K-Drama Aficionado · 8.6 — XHS-CN-Chinese audience for K-content-watching young adults.
15.3 Bottom-Resonance Cells (≤ 5.0)
Six cells under 5.0. Do not deploy these template × persona combos. They were authored for other personas; force-fitting wastes spend.
T5 (XHS Bridge) × Halal Family Provider · 3.0 — XHS is not a halal-family-provider channel. Use T4 (Merdeka) or T8 (CNY Math) instead.
T6 (Spice-Ladder Dare) × Halal Family Provider · 2.8 — Spice-flex challenge is anti-kid-friendly. Don't.
T3 (Chimaek) × Halal Family Provider · 3.6 — Late-night daypart is wrong occasion for this persona.
T7 (Live-Event) × Halal Family Provider · 3.8 — Concert tie-in is wrong occasion.
T2 (Ambassador) × Halal Family Provider · 4.2 — Halal-family-provider trust signal is JAKIM cert + product, not celebrity face.
T8 (Reunion Math) × K-Drama Aficionado · 4.8 — CNY family math doesn't resonate with solo / dyad drama-bingers.
15.4 Projected SKU / Daypart Verdicts
🏆 Flagship: Seoul Chimaek Hour daypart (T3)
9.4 / 10
Top resonance for Social Gatherer (9.4) + K-Drama (9.2). Whitespace claim no MY fried-chicken brand can match. Creates a new revenue stream, not just an ad angle. The flagship for the pitch.
⚠ Operational blocker (added 2026-05-28): All KyoChon outlets currently close 10:00 PM. The 9.4 resonance assumes 9pm-2am is deployable; it isn't without one of three pivots: (a) GrabFood delivery-only, (b) mall-negotiated extended-hours flagships, (c) 7pm-10pm "After-Mall Chimaek" re-time. Resonance score will degrade under (a) and (c). Decision required pre-launch.
🥈 Secondary: T2 + T7 — Byeon Woo-Seok Cultural Arc
8.5 / 10
Two templates compounding across June + November. Talent already paid by HQ. CTR-leader archetype (S4). Dependency: MY territorial rights clearance.
🥉 Tertiary: T1 — Fried-to-Order Reframe
7.8 / 10
High row-mean across all 4 personas — least risky template, paid-acquisition workhorse. Directly disarms biggest brand complaint (wait time). Gating dependency: outlet ops audit complete on 1U + MyTown.
15.5 9-Point Delta vs Step-2 ICP
Confirm + new + risk signals surfaced by the synth pass against the original ICP catalogue. 3 confirmed (alpha-of-record holds), 3 new signals (refresh ICP doc), 3 risks (must mitigate before deploy).
Confirmed (ICP doc is right)
K-Drama Aficionado as primary acquisition target — every top-resonance cell features them
Halal Family Provider as scale-and-frequency target — T4 (Merdeka) + T8 (CNY) both score ≥8.9 for them
Social Gatherer as moment-driven burst — T3 (Chimaek) + T7 (Live Event) both score ≥9.1 for them
New signal (refresh ICP doc)
Urban Office Worker has higher T-mean than expected (7.4) — they're not just gifting-buyers, they're a workday-lunch retention target. Tighten the dosirak SKU positioning for them.
Halal Family Provider is firmly anti-K-pop / anti-spice-challenge. Run separate creative lane — no Byeon Woo-Seok face in their ad set.
Social Gatherer is the highest-Σ "moment" persona — pre-book talent / live-event budget against this segment, not against general brand awareness.
Risk (must mitigate)
R1. Fast-synth verdict (Claude-projected) ≠ behavioural-grounded MiroFish OASIS verdict. Re-run Step 5 when backend restored; do not commit >$50/day per cell on synth-only verdicts.
R2. Halal Family Provider has all 5 bottom cells in T3/T6/T7/T2 — if any of those templates leak into their FB feed via mis-targeting, brand-trust damage compounds (the persona is risk-averse).
R3. The 24-alpha set is below the >50-alpha threshold for full creative coverage. Synth verdict optimism may not survive a real-data Meta Ad Library run.
15.6 Chain-Updates Pushed Back Upstream
To 03-icp-catalog: add explicit "kid-friendly" annotation to Halal Family Provider · annotate Urban Office Worker dosirak lunch occasion as primary, not secondary
To 10-semantic-particles: expand variant set with 10 derivative fills per S3+S5 archetype (so total alphas pass the ≥50 QA threshold)
To 11-pattern-decoder: annotate T3 as flagship explicitly (already done in this artefact via verdict)
To 23-competitive-landscape: Highlight 4Fingers Kimchick Szn as the seasonal-drop format to counter-clone — T6 spice-ladder is the clone vehicle
To 24-positioning-visual: commission chimaek-hour late-night photography refs alongside heritage refs
15.7 Composite Score
Projected Composite
8.6 / 10
Greenlight (provisional). Real OASIS re-run required to confirm. Top-SKU resonance = 9.4 (chimaek daypart × social-gatherer) clears the ≥7.5 threshold. Halal-family-provider weakness (5.4 column-mean) drags composite — requires separate creative lane to lift to ≥7.0.
15.8 Source Ledger
Resonance scoring methodology: Claude orchestrator (this session) projected resonance per cell by combining: (1) persona vocabulary match against template skeleton, (2) persona occasion fit against template's offer, (3) persona top-3 objection coverage by template, (4) persona channel mix match with template's lead channel. Each dimension scored 0-10, simple mean.
Not behavioural-grounded: This is not the MiroFish OASIS verdict. Real OASIS would route 108+ AI agents through 72 simulated hours, surfacing emergent second-order responses. Real-run pending backend restart.
Composite: mean of top-1 template per persona, averaged across the 4 personas = 8.65 ≈ 8.6. Provisional.
Real-OASIS dependency: MiroFish backend at http://localhost:5001/api/projects · POST /api/simulation/create → /api/simulation/prepare → /api/simulation/run_parallel → /api/report/generate → JSONL action logs. Re-fire scripts/fire.sh --step 5 --resume-from 5 when backend healthy.