Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Retrospective sau khi campaign ket thuc: scorecard so lieu vs target, what worked / what didn't / why co evidence, surprises, bai hoc cap nhat vao playbook + Brand Hub, action items co owner. Dung khi nhac 'retro campaign', 'retrospective', 'post-mortem', 'tong ket chien dich', 'bai hoc campaign'.
.claude/skills/minhnv0807-63-campaign-retrospective/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 137% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 71% | 0% |
> Retrospective khong phai de blame ai — ma de team va system tot hon sau moi campaign. Insight tot nhat den tu failure, nhung can tam the an toan de noi that. Retro khong co action = lang phi thoi gian: moi bai hoc phai dan den 1 thay doi cu the trong playbook, Brand Hub hoac quy trinh.
Hoi toi da 4 cau:
07-bao-cao-marketing, dung lam dau vao.)| KPI | Target | Actual | % Achieved | vs Campaign truoc | |-----|--------|--------|-----------|-------------------| | Revenue | | | | | | Leads | | | | | | CPL | | | | | | ROAS | | | | | | Spend | | | | | | Timeline (dung tien do?) | | | | | | Budget (dung ngan sach?) | | | | |
Overall verdict: Success (> 90% KPI) / Mixed (60-90%) / Miss (< 60%).
| # | What worked | Evidence (so lieu) | Why it worked | Ap dung lan sau | |---|-------------|--------------------|---------------|-----------------| | 1 | | | | | | 2 | | | | | | 3 | | | | |
Winning patterns can goi ten ro:
| # | What failed | Evidence | Root cause | Decision: Stop / Fix / Test lai | |---|-------------|----------|-----------|--------------------------------| | 1 | | | | | | 2 | | | | |
Root cause phai dao den nguyen nhan goc — "ads kem" chua phai root cause; "audience trung lap voi campaign cu → frequency cao → CTR giam" moi la root cause.
| Hang muc | Danh gia | Van de gap | Cai thien | |----------|----------|------------|-----------| | Brief quality | | | | | Timeline adherence | | | | | Communication trong team | | | | | Review/approval process | | | | | Launch day execution | | | | | Data/tracking setup | | | |
Biggest process bottleneck: 1 dieu can cai thien nhat lan sau.
Ghi nhan dong gop cu the: Ten] — contribution cu the da tac dong tich cuc den ket qua]. Ghi nhan cong khai, gop y rieng tu.
| Tai lieu can update | Thay doi gi | Owner | Deadline | |---------------------|-------------|-------|----------| | Launch playbook (60-launch-playbook) | Them/sua step: X] | Leader | | | Campaign brief template | Dieu chinh section: Y] | Leader | | | Brand Hub — Customer Insight | Them insight moi tu campaign: Z] | | | | Brand Hub — KPI Framework / benchmark noi bo | Cap nhat benchmark thuc te (CPL, CVR dat duoc) | | | | Winning creative library | Luu creative winner + ly do | | |
Moi action item phai co owner va deadline — khong co "ca team cung lam".
Ten file: campaign-retro-[ten-campaign]-[YYYYMMDD].md — gom 9 muc theo quy trinh tren: Scorecard → What worked → What didn't → Surprises → Process review → Team callouts → Playbook/Brand Hub updates → Input campaign sau → Action items co owner.
07-bao-cao-marketing: report final la dau vao data cho scorecard.60-launch-playbook: bai hoc execution update truc tiep vao playbook.02-brief-chien-dich: muc "Bai hoc tu campaign truoc" trong brief moi lay tu retro nay.34-ai-marketing-os: insight moi + benchmark thuc te cap nhat vao Brand Hub.65-team-performance-review: team callouts va observations la input cho danh gia ca nhan cuoi ky.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 21,337 | 29,788 | +40% | 1 | 1 | 0% | 3,115 | 5,634 | +81% | 0 | 0 | — |
case-02 | fail→fail | 27,829 | 28,662 | +3% | 1 | 1 | 0% | 3,577 | 6,377 | +78% | 0 | 0 | — |
case-03 | fail→fail | 27,965 | 32,859 | +18% | 1 | 1 | 0% | 3,322 | 6,808 | +105% | 0 | 0 | — |
case-04 | pass→pass | 26,673 | 29,150 | +9% | 1 | 1 | 0% | 4,229 | 5,793 | +37% | 0 | 0 | — |
case-05 | pass→pass | 23,286 | 23,195 | -0% | 1 | 1 | 0% | 3,125 | 5,434 | +74% | 0 | 0 | — |
case-01 | fail→pass | 27,785 | 30,148 | +9% | 1 | 1 | 0% | 3,771 | 5,804 | +54% | 0 | 0 | — |
case-07 | pass→pass | 17,783 | 10,477 | -41% | 1 | 1 | 0% | 2,121 | 3,755 | +77% | 0 | 0 | — |
case-08 | pass→pass | 15,498 | 11,014 | -29% | 1 | 1 | 0% | 1,699 | 2,974 | +75% | 0 | 0 | — |
case-09 | pass→pass | 11,447 | 9,440 | -18% | 1 | 1 | 0% | 1,874 | 3,369 | +80% | 0 | 0 | — |
case-10 | fail→pass | 11,628 | 20,537 | +77% | 1 | 1 | 0% | 1,763 | 4,175 | +137% | 0 | 0 | — |
case-11 | pass→pass | 16,042 | 20,910 | +30% | 1 | 1 | 0% | 1,852 | 3,594 | +94% | 0 | 0 | — |
case-12 | pass→pass | 19,398 | 20,136 | +4% | 1 | 1 | 0% | 2,276 | 4,229 | +86% | 0 | 0 | — |
case-13 | pass→pass | 30,752 | 20,877 | -32% | 1 | 1 | 0% | 3,134 | 4,230 | +35% | 0 | 0 | — |
case-14 | pass→pass | 16,613 | 21,663 | +30% | 1 | 1 | 0% | 2,464 | 4,544 | +84% | 0 | 0 | — |
case-15 | fail→pass | 21,943 | 21,334 | -3% | 1 | 1 | 0% | 2,646 | 4,855 | +83% | 0 | 0 | — |
case-16 | fail→pass | 20,691 | 11,239 | -46% | 1 | 1 | 0% | 2,338 | 3,697 | +58% | 0 | 0 | — |
case-17 | pass→pass | 14,338 | 19,860 | +39% | 1 | 1 | 0% | 2,225 | 4,029 | +81% | 0 | 0 | — |
case-18 | pass→pass | 21,504 | 24,034 | +12% | 1 | 1 | 0% | 2,420 | 4,885 | +102% | 0 | 0 | — |
case-19 | fail→pass | 7,597 | 2,744 | -64% | 1 | 1 | 0% | 1,415 | 2,422 | +71% | 0 | 0 | — |
case-20 | fail→pass | 21,015 | 16,801 | -20% | 1 | 1 | 0% | 2,421 | 4,544 | +88% | 0 | 0 | — |
case-21 | fail→pass | 23,345 | 13,691 | -41% | 1 | 1 | 0% | 3,015 | 3,152 | +5% | 0 | 0 | — |
case-22 | fail→pass | 20,991 | 16,336 | -22% | 1 | 1 | 0% | 2,724 | 4,678 | +72% | 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 +36 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.