Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Nghien cuu audience cho ads truoc khi setup campaign: xac dinh tep target, map interest/behavior theo kenh, danh gia do rong tep, phan tang cold/warm/hot, va lap gia thuyet targeting de test. Kich hoat khi user nhac 'nghien cuu audience', 'tep target', 'interest targeting', 'audience cho campaign', 'nham dung tep'.
.claude/skills/minhnv0807-51-audience-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 94% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 137% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 104% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 122% | 0% |
> Targeting sai = dot tien du copy hay, creative dep. Skill nay la buoc dau tien cua chuoi performance: 51 → 10-tinh-kpi-nguoc → 54-media-plan → 53-tracking-setup → 52-account-structure. Neu chua co insight khach hang, goi 09-insight-khach-hang truoc.
Doc .agents/product-marketing-context.md va output cua 09-insight-khach-hang neu co. Neu thieu, hoi toi da 4 cau:
10-tinh-kpi-nguoc)Tong hop tu data khach hien tai (CRM, page insight, don hang) + insight tu 09-insight-khach-hang:
| Hang muc | Chi tiet | |----------|----------| | Tuoi | khoang tuoi chinh + khoang phu] | | Gioi tinh | ti le nam/nu theo data thuc] | | Dia ly | tinh/thanh uu tien — theo kha nang giao hang/phuc vu] | | Thu nhap uoc tinh | khop voi muc gia san pham] | | Nghe nghiep / vai tro | nhom chinh] | | Thiet bi chinh | Mobile / Desktop / Ca hai |
Chi lam cho kenh se chay. Moi kenh output 1 block san setup:
| Tier | Dinh nghia | Signal | Nguon | |------|-----------|--------|-------| | Hot | Vao LP, add to cart, mo form, inbox chua chot | Pixel events, inbox | Meta, Google, CRM | | Warm | Xem video >50%, engage page/post, click link, follow | Engagement custom audience | Meta, TikTok | | Cold | Chua biet brand | Interest / behavior / broad / LAL | Tat ca kenh |
Quy tac: cold va warm/hot phai nam o campaign khac nhau, message khac nhau (xem 56-retargeting-plan). Luon exclude warm/hot khoi campaign cold de data sach.
| Seed | Size toi thieu | LAL % | Platform | Muc tieu | |------|---------------|-------|----------|---------| | Khach da mua | >=100 | 1-3% | Meta | Tim nguoi giong khach tot nhat | | Lead chat luong | >=500 | 1-5% | Meta | Scale lead gen | | Video viewer 75% | >=1000 | Broad LAL | TikTok | Scale awareness | | Email list | >=300 | 1-5% | Meta | Mo rong tu CRM |
Seed cang chat (khach mua > lead > viewer) thi LAL cang chat luong. Khong tao LAL tu seed duoi nguong size.
Moi gia thuyet 1 dong, dang "Neu target X] thi metric] se Y] vi Z]" — chuyen cho 19-ab-test-setup va 52-account-structure de len ad set:
| # | Gia thuyet | Bien test | Metric do | Uu tien | |---|-----------|-----------|-----------|---------| | 1 | Broad + creative manh re hon interest | Broad vs Interest | CPL | Cao | | 2 | Interest nhom A] sat pain hon nhom B] | Interest A vs B | CPL, CTR | Cao | | 3 | LAL 1% khach mua chat hon LAL 3% | LAL % | CPL, close rate | Trung binh | | 4 | Age segment X] chuyen doi tot hon | Age split | CPL | Thap |
Toi da 3-5 ad set test cung luc — nhieu hon thi budget loang, khong du data ket luan.
Ten file: audience-research-[ten-san-pham]-[YYYYMMDD].md
markdown# Audience Research — [San pham] Ngay: [YYYY-MM-DD] · Kenh: [Meta/TikTok/...] · Objective: [Lead/Conversion/...] ## 1. Core audience profile | Tuoi | Gioi | Dia ly | Thu nhap | Nghe nghiep | Thiet bi | ## 2. Psychographic & behavior - Pain points: [3-5, ngon ngu khach hang] - Motivation: [muon gi / tranh gi] - Online behavior: [nen tang, gio peak, format ua thich] ## 3. Targeting settings theo kenh ### Meta: [age/gender/location/interests theo nhom/behaviors/exclusions] ### TikTok: [...] · Google: [...] · Zalo: [...] ## 4. Audience tiers | Tier | Audience | Signal | Campaign su dung | ## 5. Lookalike seeds | Seed | Size | LAL % | Platform | Muc tieu | ## 6. Gia thuyet targeting (chuyen cho A/B test) | # | Gia thuyet | Bien test | Metric | Uu tien | ## 7. Ghi chu do rong tep - Audience size uoc tinh tung ad set: [so] — cold nen o 100K-500K - Overlap can tranh: [cac tep de trung nhau]
55-scaling-ads). Tep thua → ghi ly do, khong test lai y het.09-insight-khach-hang: chay truoc — cung insight, pain, ngon ngu khach hang.08-nghien-cuu-doi-thu: xem doi thu dang target ai, angle nao (Ads Library).10-tinh-kpi-nguoc: tinh CPL max / budget truoc khi len plan.54-media-plan: dung audience nay de phan bo kenh va budget.52-account-structure: chuyen targeting settings thanh cau truc ad set.56-retargeting-plan: chi tiet phan tang warm/hot va message theo tang.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 41,934 | 35,135 | -16% | 1 | 1 | 0% | 6,607 | 7,890 | +19% | 0 | 0 | — |
case-01 | fail→fail | 36,137 | 33,340 | -8% | 1 | 1 | 0% | 6,197 | 7,107 | +15% | 0 | 0 | — |
case-02 | fail→fail | 31,551 | 35,748 | +13% | 1 | 1 | 0% | 5,019 | 8,154 | +62% | 0 | 0 | — |
case-04 | fail→fail | 15,839 | 18,109 | +14% | 1 | 1 | 0% | 2,394 | 5,281 | +121% | 0 | 0 | — |
case-05 | pass→pass | 10,093 | 9,691 | -4% | 1 | 1 | 0% | 1,661 | 3,905 | +135% | 0 | 0 | — |
case-06 | fail→pass | 15,284 | 19,042 | +25% | 1 | 1 | 0% | 2,401 | 4,669 | +94% | 0 | 0 | — |
case-07 | pass→pass | 22,356 | 17,227 | -23% | 1 | 1 | 0% | 2,546 | 5,162 | +103% | 0 | 0 | — |
case-08 | fail→pass | 25,295 | 15,565 | -38% | 1 | 1 | 0% | 2,570 | 4,750 | +85% | 0 | 0 | — |
case-09 | pass→pass | 14,635 | 18,219 | +24% | 1 | 1 | 0% | 1,571 | 4,480 | +185% | 0 | 0 | — |
case-10 | pass→pass | 20,252 | 20,568 | +2% | 1 | 1 | 0% | 2,144 | 4,803 | +124% | 0 | 0 | — |
case-11 | pass→pass | 21,491 | 19,989 | -7% | 1 | 1 | 0% | 2,536 | 5,411 | +113% | 0 | 0 | — |
case-12 | fail→pass | 13,833 | 8,817 | -36% | 1 | 1 | 0% | 1,297 | 3,077 | +137% | 0 | 0 | — |
case-13 | pass→pass | 27,198 | 18,745 | -31% | 1 | 1 | 0% | 3,326 | 5,334 | +60% | 0 | 0 | — |
case-14 | fail→pass | 18,645 | 27,381 | +47% | 1 | 1 | 0% | 2,931 | 5,976 | +104% | 0 | 0 | — |
case-15 | fail→pass | 14,560 | 10,954 | -25% | 1 | 1 | 0% | 1,531 | 3,400 | +122% | 0 | 0 | — |
case-16 | pass→pass | 14,895 | 19,464 | +31% | 1 | 1 | 0% | 2,220 | 4,553 | +105% | 0 | 0 | — |
case-17 | pass→pass | 14,207 | 21,268 | +50% | 1 | 1 | 0% | 2,258 | 4,631 | +105% | 0 | 0 | — |
case-18 | pass→pass | 17,923 | 18,189 | +1% | 1 | 1 | 0% | 2,084 | 4,754 | +128% | 0 | 0 | — |
case-19 | fail→fail | 22,891 | 21,907 | -4% | 1 | 1 | 0% | 3,368 | 5,133 | +52% | 0 | 0 | — |
case-20 | fail→fail | 22,909 | 16,362 | -29% | 1 | 1 | 0% | 2,515 | 5,060 | +101% | 0 | 0 | — |
case-21 | fail→fail | 29,666 | 24,257 | -18% | 1 | 1 | 0% | 5,678 | 7,367 | +30% | 0 | 0 | — |
case-22 | fail→fail | 26,674 | 25,303 | -5% | 1 | 1 | 0% | 4,361 | 5,961 | +37% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +23 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.