HK Agency

The winning BI-SU ad (#340) is fatiguing after a month at the top, so the meeting builds its successor by going deeper into the narrative that actually won — rarity and luxury, not science. The audience data is unusually clean: women 35–45, Instagram Feed + Stories, 9:16, iPhone, 42 conversions — profiled as a well-off ~43-year-old who is skeptical of scientific claims and buys to feel like a discerning collector, comfortable at $100–400+ on premium skincare. Three new angles get built to replace #340: Rarity Index (vs saffron, diamonds, royal jelly), Price Comparison (gram-for-gram vs silver and gold) and Scarcity (limited "harvest drops", mirroring the brand's Japanese playbook). Visuals move from "20s lifestyle" to a Vogue-magazine aesthetic — older, more diverse avatars, luxury interiors instead of subways — plus luxury unboxing UGC and Kyoto / Ginza store brand-story ads for credibility. Second priority: BC16 is selling after its LP tweak → ship more BC16 ads immediately with the same luxury/scarcity angles and ice-themed summer visuals.

Strategy meeting that trades volume for quality: stop pumping out video ads that don't convert and put the hours into deep research first (Reddit, Amazon reviews, customer feedback) so each creative names a specific problem — justified by a Meta engineering post revealing a shift to "hierarchical interest representation" (the algorithm now models users at "drinks black coffee at 9am" resolution, not "likes coffee"). Biggest new bet: give BC16 (the edible beauty stick) its first paid budget — it sells organically on zero ad spend, so duplicate a winning campaign onto its LP, build a 3-flavor mixed-pack LP (nutrition facts ×3 + subscription, Japanese text removed) and adapt the winning 340 / 343 creative onto it. Team roles rebalance: Han moves off video production onto influencer outreach + image-ad animation to unload Kim. Blocking check before any of it launches: verify the subscription app actually talks to the custom Shopify theme — Hide places a real test order, team + app support confirm auto-billing and order creation.

Execution session on the BI-SU push. Core discipline installed: test ONE variable at a time (proven visual + new hook — e.g. the "184" dog + a new script) so wins/losses are attributable. Launch Campaign 7 (C7) at half the main budget to test BI-SU creatives in isolation while current ads stay in C36 — C7 gets BC16 luxury images, a new summer pineapple / Vitamin-C angle, and a rose+cranberry (red-scheme) variation of the winning pineapple ad. Ship the new BI-SU LP "Asia L32" once the off-center product image is replaced and the subscription checkout is test-ordered. Influencer plan: scale BI-SU outreach (75→8→4 is working), keep Miranest small + selective; decide on the new content platform ($500/mo, a $60 proposal) after clarifying usage/partnership-ad rights. New creative guardrails: GLP-1 reference videos must be viral (8k+ likes, not ~500) before templating, and all visuals must match the US season (summer).

A new angle with real edge: an ingestible supplement that prevents the facial sagging caused by GLP-1 weight-loss drugs (Ozempic). Competitors (Sinatra, IMA) sell weight loss; nobody sells the side-effect fix — unique market position. Ship it as a micro-budget Meta test (45–54 women, 9:16, stats-image ad + new LP). Two process upgrades: (1) lock influencer partnership rights (unlimited ad use) + scope BEFORE quoting the client — a recent $600 approval nearly doubled when the creator later added $300 for ad rights; (2) an AI production stack — Claude for strategy, Codex for heavy coding (token-saving), Kimi as a cheap model, Obsidian for local notes — used to build LPs and auto-QA the checkout backend. Miranest data is healthy again (CPO recovered from a >$38 spike; video ≈ image; FB/IG Feed win; core = 45–54 women on iPhone).

Germany is paused. Miranest DE ads are switched off for poor performance and the budget moves to the US, which got a +20% increase — the month-end verdict from Jul 17 landed early, and it landed against Germany. The US earned it: performance nearly doubled in the last 30 days on the same budget. But that's exactly the risk — it all rides on a single winning creative, so the mandate is diversify now (new angles, videos + images) before fatigue hits. Process upgrade: a new AI-powered sheet becomes the center of gravity — it auto-pulls winning ads from Meta, analyzes them (avatar · problem · desire), returns full data (CPO, retention, demographics), and generates scripts, building a searchable library of proven strategies. New standing rule: every creative idea must arrive with supporting data.

Friday readout — and the numbers are bad. Miranest USA is falling: conversions 44 → 19 week-over-week, with Campaign 32 burned out (Frequency 6 = audience saturation). Worse: Miranest Germany is at risk of being pulled by the client if performance doesn't improve by month-end — now the top priority. Two hard outputs: (1) a non-negotiable creative rule — every new ad must show a dog's face in the first 3 seconds (every winner has it; ads without it fail); (2) a 3-pipeline creative test to find which process actually produces winners — Hide (mimic Meta Ad-Library winners), Nine (Reddit pain-points → hypothesis), Han (AI-generated angles + BI-SU library for UGC reference) — all tracked ideation→launch in one template with auto-synced Meta data. Also: AI influencer sourcing failed (8 AI lists → 2 qualified) — manual sourcing stays.

Germany converts badly because German shoppers don't trust us — and the site gives them no reason to. Missing trust signals: an Impressum (legally required company details), an active social presence, Klarna (>60% of German online shoppers), and reliable shipping. The content actively hurts too: "superfood" reads as meaningless, "93% of 30k" is dismissed without a linked source, and 450 US reviews with non-German names aren't credible. The new strategy: stop guessing — "biocode" the German competitor "Mamling" (clean site, ~$200k/month): copy its structure, design, and tone wholesale, then swap in our product with compliant copy — the same play that worked in the US with "Woofies". The gate is technical: our Shopify theme is too restrictive, so this hinges on Hydrogen API or a third-party app. Also: drop Sharkline → manual IG outreach.

Two big pivots + a creative-method reset. (1) Kill the minor-ad-variation cycle — Facebook treats small tweaks of a winner as the same creative (diminishing returns), so adopt structured frameworks (Hide's 6-component video structure + Han's winning-ad research) to build entirely new US ads with fresh hooks/scripts, tested on winners N164 / N24. (2) Completely rebuild the German market to fit local behavior — a new landing page from scratch (not a translation), one-time purchase as the default, native payments (Klarna / bank transfer), hard-data messaging, drop "before/after" (reads as a scam in DE), 14-day no-questions returns. (3) Influencer pivot to <50k high-engagement micro-creators with full creative freedom + rigorous KPI tracking. Everything is now KPI-tracked and data-justified; team presents Fri Jul 17, 3pm.

This is a research week, not an output week — every team member presents findings Friday July 17 (after 3pm) against a fixed structure: what did you find · what do you think · how do we apply it to Miranest/Bisu. Strategic thrust: diversify content to de-risk the single-ad reliance — real influencers → social proof (replacing the failed AI-UGC experiments) and animations → brand storytelling. Concrete quotas: Huma finds 10 free influencers (Miranest), Kim finds 15 free + 3 paid (Bisu) and builds a "rarest beauty ingredients" landing page, and the two of them build an AI tool for bulk influencer discovery. New research/content engine adopted: NotebookLM (source-grounded, no hallucination).

Reset the influencer program and pivot the whole shop from mass output → research-first quality. Huma re-contacts ghosted influencers under the alias "Hannah" — apologize (Tina left), promise timely comms, offer free jelly sticks to rebuild trust. New rule: prioritize engagement (comment activity) over follower count — ~$80/package JP shipping makes mass-sending unsustainable. Isabella = top ambassador target (~500 comments/post, already uses Miranest) → propose a paid $300–$500/mo, 3–6 month deal to license her content for ads + landing pages. All new creative must be built on proven research (competitor breakdowns), not raw ideas. Germany is bleeding (1 sale in 3 days) → test three new 10% lookalikes (Pixel / Purchase-180 / Shopify list), research German shopping habits, and lean into educational content (out-converting before/after in the US).

Fix Miranest Germany's poor CPO — root cause is self-competition (our agency + PSJN running similar ads on the shared Miranest Global page/pixel, inflating auction prices). Stop current campaigns and test two new DE campaigns, 25k JPY each, 7-day read: (1) a 1% purchase LLA from a combined US+DE customer list, (2) broad targeting on a brand-new Facebook page to reset delivery. Upgrade influencer vetting (judge comment purchase intent, not vanity engagement). Shift creative to educational animation + storytelling ("Lazial" / "Swiftlet Revolution" style). New hire Mao (17, Japan) joins to learn digital marketing.

Shift ad strategy to quality over quantity — 80% deep research (viral hooks + proven templates), 15% ideation, 5% AI-assisted editing — before mass production. Influencer outreach is the top priority (Thu client presentation). For Germany: pivot messaging from "Swallow's Nest" → "superfood", target the German Spitz breed (Pomeranian before/afters as proxy), and replicate proven US templates F2/F45.

Walkthrough of the new HK Agency platform — the team's central hub for performance tracking, to-do lists, meeting summaries and idea sharing. Judge daily performance on Meta CPO (🟢 <¥25k · 🟡 ¥25–35k · 🔴 >¥35k); for video ads chase a >25% hook rate (viewers held through the first 3s). Weekly rhythm: build a personal to-do list + check meeting summaries for team action items.

Revenue dropped ~75% (single-client dependence) → agency now at a loss. Two moves: (1) fix underperforming AI ads by mimicking winning UGC frame-for-frame, and double ad spend next month to hit break-even; (2) de-risk by launching a proprietary AI-enhanced digital product/service — shift from service agency → product company.

Past influencer campaigns made posts but didn't convert. Shift: treat influencers as content creators for ads, not social posters. Two tiers — free (volume, commission) + paid (quality, flat fee + product) — with strict vetting. $3,000/mo test budget across Bisu/Miranest.