Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Setup he thong tracking truoc khi chay ads: Pixel + Conversion API, GA4, UTM convention, event mapping, offline conversion, va checklist verify xanh truoc khi len camp. Quy tac cung: khong chay ads khi tracking chua verify. Kich hoat khi user nhac 'setup tracking', 'pixel', 'UTM', 'conversion API', 'CAPI', 'GA4', 'do luong campaign'.
.claude/skills/minhnv0807-53-tracking-setup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 103% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 107% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 79% | 0% |
> Ads khong co tracking = lai xe bit mat. QUY TAC CUNG: khong chay ads khi tracking chua verify xanh — chay la lang phi 100%. Setup dung tu dau; sua tracking giua campaign lam hong data lich su.
Hoi toi da 4 cau:
Meta Ads — Pixel + CAPI (bat buoc ca hai):
TikTok Ads:
Google Ads:
Zalo Ads: cai Zalo Pixel, setup conversion, verify. YouTube Ads: dung chung conversion tracking cua Google Ads.
https://domain.com/lp?utm_source=[source]&utm_medium=[medium]&utm_campaign=[campaign]&utm_content=[ad_name]| Parameter | Quy uoc | Vi du | |-----------|---------|-------| | utm_source | facebook / tiktok / google / zalo / email | facebook | | utm_medium | paid_social / paid_search / display / email / organic | paid_social | | utm_campaign | [product]_[audience]_[MMYY] | course-a_cold_0726 | | utm_content | [creative-type]_[hook-type]_[version] — khop ten ad | vid_pain_v1 |
Quy tac: viet thuong, khong dau, khong khoang trang. utm_content phai trung ten ad trong 52-account-structure de doi chieu 2 chieu Ads Manager ↔ GA4.
| Event GA4 | Trigger | Map voi platform event | |-----------|---------|------------------------| | page_view | Tat ca trang | PageView (Meta), ViewContent (TikTok) | | view_item | Xem trang san pham/offer | ViewContent (Meta) | | generate_lead | Submit form / click inbox | Lead (Meta), SubmitForm (TikTok), Conversion (Google) | | purchase | Thanh toan thanh cong | Purchase (Meta/TikTok/Google) |
Moi event chi fire 1 lan cho 1 hanh dong. Dat conversion chinh trong GA4 = event khop objective ads.
Ap dung khi chot don qua inbox / goi dien / tai quay (pho bien o VN):
ID | Ngay vao | Ten | SDT | Kenh nguon | Loai tuong tac (inbox/form/goi) | Trang thai (moi/qualified/chot/khong chot) | Ly do khong chot | Doanh thu neu chot | Sales phu trach. Hoi moi khach "biet minh qua kenh nao?" ngay cau dau.Neu website chi phuc vu khach trong nuoc, phan nay tuan theo Nghi dinh 13/2023/ND-CP ve bao ve du lieu ca nhan: phai co thong bao ro rang truoc khi thu thap, va nguoi dung phai dong y — khong duoc mac dinh tick san.
Neu co bat ky traffic tu EU/UK hoac California (khach Viet kieu, ban hang xuyen bien gioi, du hoc, xuat khau), nhung yeu cau sau la BAT BUOC — thieu la rui ro phat tien va mat quyen dung remarketing audience:
| Vung | Yeu cau | Hau qua neu thieu | |------|---------|-------------------| | EU / UK | CMP duoc chung nhan + Google consent mode v2 (4 tin hieu: ad_storage, analytics_storage, ad_user_data, ad_personalization, mac dinh denied truoc khi user tra loi banner) | Meta yeu cau CMP cho traffic EU; Google mat personalized ads va audience feature | | California | Nut tu choi ban/chia se du lieu, ton trong tin hieu Global Privacy Control, bat Limited Data Use tren Meta cho traffic California | Vi pham CCPA/CPRA | | Toan cau | Tag khong duoc ban truoc khi co dong y; luu log dong y (ai, chon gi, luc nao, phien ban chinh sach nao) | Khong chung minh duoc tuan thu khi bi kiem tra |
Danh sach email/SDT upload lam Custom Audience hoac Customer Match cung phai co dong y tuong ung — xem 14-email-marketing.
Ghi vao file ket qua: dung CMP nao, chan vung nao, ai verify.
Tat ca phai PASS — thieu 1 muc = KHONG LAUNCH:
Daily (2 phut): Events Manager 24h co event deu khong; Google Ads conversions dang ghi; GA4 co session tu paid. Weekly: CAPI match rate > 70%; co duplicate event khong; so lead 3 nguon (Ads Manager / GA4 / Lead Log) lech <20%; UTM co link nao thieu/sai.
| Van de | Nguyen nhan thuong gap | Cach fix | |--------|------------------------|----------| | Pixel khong fire | Cai sai vi tri / thieu tren LP | Check GTM preview, hard-code lai header | | Event duplicate | Cai ca GTM lan hard-code | Chon 1 cach, go cach kia; check event_id dedup | | CAPI match rate thap | Khong gui email/phone | Them customer data vao payload CAPI | | Conversion khong ghi | Sai ten event / sai trang trigger | Check chinh ta event, trigger dung thank you page | | GA4 lech xa Ads Manager | Attribution window khac nhau + UTM thieu | Chuan hoa UTM, so sanh cung window | | Lead nhieu nhung khong ro nguon | Quen hoi kenh, khong ghi Lead Log | Bat buoc cau hoi nguon trong script sales |
Ten file: tracking-setup-[ten-san-pham]-[YYYYMMDD].md
markdown# Tracking Setup — [San pham/Campaign] Platform: [danh sach] · Website: [nen tang] · Objective: [Lead/Purchase] ## 1. Checklist setup theo platform [bang trang thai tung muc] ## 2. UTM convention [bang parameter + vi du cua project] ## 3. Event mapping [bang GA4 ↔ platform events] ## 4. Offline conversion & Lead Log [cach ghi + lich upload] ## 5. Verify truoc launch [checklist xanh — ngay verify + nguoi verify] ## 6. Lich check daily/weekly [ai lam, luc nao]
52-account-structure: setup campaign SAU khi tracking xanh; utm_content khop ten ad.54-media-plan: plan xac dinh objective → quyet dinh event can track.21-audit-ads-performance: audit tracking dinh ky (EMQ, duplicate, match rate).11-thiet-lap-kenh: setup kenh va tai san (page, OA, website) truoc tracking.13-phan-tich-du-lieu: doc data sach tu he thong tracking nay.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 35,266 | 37,417 | +6% | 1 | 1 | 0% | 5,784 | 8,142 | +41% | 0 | 0 | — |
case-02 | fail→pass | 37,850 | 30,780 | -19% | 1 | 1 | 0% | 5,225 | 6,793 | +30% | 0 | 0 | — |
case-03 | fail→fail | 36,739 | 42,584 | +16% | 1 | 1 | 0% | 5,781 | 8,634 | +49% | 0 | 0 | — |
case-04 | pass→pass | 20,997 | 29,541 | +41% | 1 | 1 | 0% | 3,494 | 6,485 | +86% | 0 | 0 | — |
case-05 | pass→pass | 18,866 | 34,960 | +85% | 1 | 1 | 0% | 2,968 | 6,838 | +130% | 0 | 0 | — |
case-06 | pass→pass | 34,606 | 28,303 | -18% | 1 | 1 | 0% | 2,713 | 6,320 | +133% | 0 | 0 | — |
case-07 | fail→pass | 28,824 | 29,344 | +2% | 1 | 1 | 0% | 3,045 | 6,376 | +109% | 0 | 0 | — |
case-08 | fail→fail | 20,851 | 29,703 | +42% | 1 | 1 | 0% | 3,275 | 6,439 | +97% | 0 | 0 | — |
case-09 | fail→pass | 20,489 | 29,400 | +43% | 1 | 1 | 0% | 3,277 | 6,657 | +103% | 0 | 0 | — |
case-10 | fail→fail | 22,380 | 20,032 | -10% | 1 | 1 | 0% | 2,468 | 5,547 | +125% | 0 | 0 | — |
case-11 | fail→fail | 29,421 | 31,445 | +7% | 1 | 1 | 0% | 3,593 | 6,646 | +85% | 0 | 0 | — |
case-12 | pass→pass | 24,240 | 27,358 | +13% | 1 | 1 | 0% | 4,025 | 7,290 | +81% | 0 | 0 | — |
case-13 | fail→fail | 23,540 | 26,870 | +14% | 1 | 1 | 0% | 3,166 | 6,187 | +95% | 0 | 0 | — |
case-14 | fail→pass | 16,294 | 19,810 | +22% | 1 | 1 | 0% | 2,965 | 6,145 | +107% | 0 | 0 | — |
case-15 | pass→pass | 19,265 | 23,461 | +22% | 1 | 1 | 0% | 3,294 | 6,592 | +100% | 0 | 0 | — |
case-16 | pass→pass | 17,626 | 24,704 | +40% | 1 | 1 | 0% | 2,911 | 5,824 | +100% | 0 | 0 | — |
case-17 | pass→pass | 21,068 | 20,405 | -3% | 1 | 1 | 0% | 3,167 | 5,830 | +84% | 0 | 0 | — |
case-18 | fail→pass | 21,475 | 12,517 | -42% | 1 | 1 | 0% | 2,549 | 4,560 | +79% | 0 | 0 | — |
case-19 | fail→fail | 19,039 | 30,958 | +63% | 1 | 1 | 0% | 2,802 | 6,714 | +140% | 0 | 0 | — |
case-20 | fail→pass | 19,428 | 23,250 | +20% | 1 | 1 | 0% | 3,596 | 6,815 | +90% | 0 | 0 | — |
case-21 | fail→pass | 16,459 | 22,358 | +36% | 1 | 1 | 0% | 2,872 | 6,427 | +124% | 0 | 0 | — |
case-22 | pass→pass | 19,568 | 30,517 | +56% | 1 | 1 | 0% | 3,108 | 6,592 | +112% | 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 +32 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.