Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Lap media plan chay ads: tinh nguoc tu doanh thu ra CPL max va budget, phan bo ngan sach theo kenh va funnel TOFU/MOFU/BOFU, budget split Testing 30/Scale 50/Retarget 15/LAL 5, timeline ramp-up theo tuan, KPI target va contingency plan. Kich hoat khi user nhac 'media plan', 'plan ads', 'ke hoach chay ads', 'phan bo ngan sach ads'.
.claude/skills/minhnv0807-54-media-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 133% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 82% | 0% |
> Plan khong co KPI cu the = khong biet thang hay thua. Moi so trong plan phai truy duoc ve doanh thu muc tieu. Goi 10-tinh-kpi-nguoc de tinh CPL max / budget truoc khi phan bo.
Doc .agents/product-marketing-context.md, output 51-audience-research va 10-tinh-kpi-nguoc neu co. Neu thieu, hoi toi da 4 cau:
10-tinh-kpi-nguoc)Doanh thu muc tieu / AOV = So don can
So don / Close rate = So lead can
Max CPL = (Doanh thu x Gross margin %) / So lead
Break-even ROAS = 1 / Gross margin %
Target ROAS = Break-even ROAS x 1.5
Budget ads can = So lead x CPL targetKet qua bat buoc co truoc khi sang buoc 2: So lead can · CPL max · CPL target · Budget ads · Target ROAS. Tinh du 3 kich ban (Pessimistic / Base / Optimistic) — lap plan bang kich ban Base.
| Kenh | % Budget | So tien/thang | Objective | KPI chinh | |------|---------|--------------|-----------|-----------| | Meta Ads | %] | so] | Lead/Conv] | CPL, CPMess | | TikTok Ads | %] | so] | Video-first] | CPL, VCR | | Google Ads | %] | so] | Search intent] | CPL, Conv rate | | Zalo / khac | %] | so] | | | | Tong | 100% | | | |
Logic phan bo mau (dieu chinh theo data rieng):
Quy tac: kenh da chung minh hieu qua nhan 60-70%; kenh moi test 15-20%; khong dan deu.
| Tang | % Budget | Audience | Creative | KPI chinh | |------|---------|----------|----------|-----------| | TOFU — Awareness/Traffic | 30-40% | Cold (interest/broad/LAL) | Hook manh, educate, entertain | CPM, CTR 0.8-2% | | MOFU — Lead gen | 35-45% | Warm (engaged 7-30 ngay) + cold da loc | Educate + offer, giai phap | CPL, lead quality | | BOFU — Conversion/Retarget | 15-25% | Hot (visit LP, cart, inbox 1-7 ngay) | Urgency, proof, testimonial | CPA, ROAS |
| Loai campaign | % Budget | Muc tieu | |---------------|---------|---------| | Testing (cold, tim winner) | 30% | Test creative + audience moi | | Scale (winner da chung minh) | 50% | Volume chinh | | Retargeting | 15% | Convert warm/hot | | Lookalike | 5% | Mo tep tu seed chat luong |
Hai goc nhin (funnel va loai campaign) phai khop nhau khi cong tong — dung bang nay lam chuan chia campaign trong 52-account-structure.
| Tuan | Giai doan | % Budget thang | Focus | Checkpoint | |------|-----------|---------------|-------|------------| | 1-2 | Testing | ~30% | Test 3-5 ad set, 2-3 creative/ad set | Dung ad set CPL > 2x target sau 3 ngay | | 3-4 | Scaling | ~50% | Scale winner +20-30%/lan, khong x2 ngay | CPL on dinh, volume tang | | 5+ | Maintaining | ~70% chay on dinh | Refresh creative, test nho lien tuc | Frequency < 2.5, ROAS >= target |
Ramp-up budget/ngay vi du (thang dau): tuan 1: 300-500K/ngay → tuan 2: 500-800K → tuan 3: 800K-1.5M → tuan 4: 1.5M-3M (theo 55-scaling-ads, chi tang khi CPL dat).
| Kenh | CPM target | CTR target | CPL target | ROAS target | |------|-----------|-----------|-----------|-------------| | Meta | theo references/benchmarks-vietnam.md] | >1.5% | tu buoc 1] | tu buoc 1] | | TikTok | | >1% | | | | Google | | >3% | | |
CPL decision rules hang tuan (vi du nganh dich vu local, CPL target 12K — chinh nguong theo AOV va CPL target cua ban):
| CPL thuc te | Frequency | Action | |-------------|-----------|--------| | ≤ 1.0x CPA target | < 2.0 | WIN — ung vien scale, kiem tra dieu kien trong 55-scaling-ads | | 1.0 - 1.5x | < 2.0 | Theo doi — toi uu, chua scale | | > 1.5x keo dai 7 ngay | 2.0-2.5 | Thay creative hoac audience | | > 3.0x | > 2.5 | TAT — ghi ly do vao log |
> Thang nay lay tu skills/vi/references/quality-gates-vn.md Gate 1 (nguon duy nhat). Truoc khi phan xu, kiem tra da spend >= 3x CPA muc tieu chua — chua du thi cho, dung tat som.
| Tinh huong | Dau hieu | Action | |-----------|---------|--------| | CPA > 3x target | Da spend >= 3x CPA target | Tat ad set (Gate 1), test creative moi | | CTR < 0.5% | Ngay 1-2 | Doi creative / hook ngay | | Frequency > 3 | Tuan 2+ | Refresh creative, mo rong tep | | Spend khong het (delivery thap) | Pace < 80% plan | Mo audience, tang bid, check reject | | CPM tang dot bien | Mua cao diem / canh tranh | Giam budget tam thoi, don sang kenh re hon |
Ten file: media-plan-[ten-san-pham]-[YYYYMMDD].md
markdown# Media Plan — [San pham/Campaign] Timeline: [tu ngay - den ngay] · Budget tong: [so] · Objective: [Lead/Conv] ## 1. Campaign overview [ten, offer, timeline, budget, KPI chinh: CPL/ROAS/Leads] ## 2. KPI tinh nguoc [bang 3 kich ban tu 10-tinh-kpi-nguoc] ## 3. Phan bo theo kenh [bang % + tien + objective + KPI] ## 4. Phan bo theo funnel [TOFU/MOFU/BOFU] ## 5. Budget split van hanh [Testing 30 / Scale 50 / Retarget 15 / LAL 5] ## 6. Timeline ramp-up [tuan, budget/ngay, focus, checkpoint] ## 7. KPI target theo kenh + decision rules ## 8. Contingency plan
10-tinh-kpi-nguoc: BAT BUOC chay truoc — cho CPL max, budget, target ROAS.51-audience-research: audience cho tung tang funnel.53-tracking-setup: verify tracking truoc khi tieu dong nao.52-account-structure: chuyen plan thanh cau truc campaign.00-ke-hoach-mkt: media plan la 1 phan cua ke hoach marketing tong.57-next-ads-plan: ky sau lap plan tu data ky nay.references/benchmarks-vietnam.md — so lieu lech thi uu tien file repo| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 31,609 | 41,501 | +31% | 1 | 1 | 0% | 6,033 | 9,834 | +63% | 0 | 0 | — |
case-02 | fail→fail | 31,469 | 28,308 | -10% | 1 | 1 | 0% | 6,278 | 8,349 | +33% | 0 | 0 | — |
case-03 | fail→pass | 32,191 | 29,222 | -9% | 1 | 1 | 0% | 6,247 | 8,560 | +37% | 0 | 0 | — |
case-04 | pass→pass | 20,339 | 18,102 | -11% | 1 | 1 | 0% | 2,980 | 5,382 | +81% | 0 | 0 | — |
case-05 | pass→pass | 17,448 | 25,028 | +43% | 1 | 1 | 0% | 3,436 | 7,031 | +105% | 0 | 0 | — |
case-06 | pass→pass | 17,795 | 21,479 | +21% | 1 | 1 | 0% | 3,305 | 6,761 | +105% | 0 | 0 | — |
case-07 | fail→fail | 19,967 | 24,418 | +22% | 1 | 1 | 0% | 3,153 | 6,672 | +112% | 0 | 0 | — |
case-08 | fail→fail | 20,675 | 17,886 | -13% | 1 | 1 | 0% | 3,164 | 5,838 | +85% | 0 | 0 | — |
case-09 | fail→fail | 20,770 | 24,554 | +18% | 1 | 1 | 0% | 3,325 | 7,096 | +113% | 0 | 0 | — |
case-10 | fail→fail | 17,270 | 24,502 | +42% | 1 | 1 | 0% | 2,939 | 7,119 | +142% | 0 | 0 | — |
case-11 | pass→fail | 13,371 | 11,555 | -14% | 1 | 1 | 0% | 2,156 | 4,681 | +117% | 0 | 0 | — |
case-12 | fail→pass | 16,260 | 14,988 | -8% | 1 | 1 | 0% | 2,860 | 5,197 | +82% | 0 | 0 | — |
case-13 | pass→fail | 16,113 | 19,046 | +18% | 1 | 1 | 0% | 2,828 | 5,997 | +112% | 0 | 0 | — |
case-14 | fail→pass | 9,419 | 6,792 | -28% | 1 | 1 | 0% | 1,649 | 3,846 | +133% | 0 | 0 | — |
case-15 | pass→pass | 19,061 | 14,501 | -24% | 1 | 1 | 0% | 2,700 | 5,077 | +88% | 0 | 0 | — |
case-16 | fail→pass | 19,079 | 14,748 | -23% | 1 | 1 | 0% | 2,769 | 5,081 | +83% | 0 | 0 | — |
case-17 | fail→pass | 14,221 | 7,883 | -45% | 1 | 1 | 0% | 2,168 | 3,936 | +82% | 0 | 0 | — |
case-18 | fail→pass | 18,345 | 18,901 | +3% | 1 | 1 | 0% | 3,304 | 6,071 | +84% | 0 | 0 | — |
case-19 | pass→pass | 14,563 | 6,319 | -57% | 1 | 1 | 0% | 2,073 | 3,632 | +75% | 0 | 0 | — |
case-20 | pass→pass | 11,543 | 13,572 | +18% | 1 | 1 | 0% | 1,912 | 4,480 | +134% | 0 | 0 | — |
case-21 | pass→pass | 14,883 | 11,079 | -26% | 1 | 1 | 0% | 2,520 | 4,380 | +74% | 0 | 0 | — |
case-22 | pass→pass | 18,517 | 16,761 | -9% | 1 | 1 | 0% | 2,905 | 5,145 | +77% | 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 +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
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.