| name | ad-analysis |
| description | Deconstructs an ad's performance data to identify why it worked or failed, extract learnings, and generate 3 test hypotheses. Use after a creative has accumulated significant spend. Trigger on: 'analyze this ad', 'why is it working', 'creative deconstruction', 'what do we learn from this result'. Produces a learning report and 3 actionable next test hypotheses. |
| metadata | {"version":"1.0.0","status":"stable","tags":["analysis","performance","creative","testing","hypotheses"],"inputs":["creative performance data (Meta metrics)","visual access or description of the creative"],"outputs":["deconstruction report + 3 actionable test hypotheses"],"depends-on":["brand-health"]} |
Ad Analysis
Before Starting
Confirm before starting:
For reference benchmarks: see references/kpi-benchmarks.md
Phase 1: Read the Data
Collect and organize key metrics:
| Metric | Value | Benchmark | Status |
|---|
| Hook Rate (video) | | 20–30% | |
| Hold Rate 50% | | 20–35% | |
| CTR (link click) | | 1–2% | |
| CPM | | 8–15€ | |
| CPA / CPL | | [target] | |
| Frequency | | < 3 | |
| Total spend | | — | |
Identify immediately: is it a winner, a learner, or a loser?
- Winner: Primary KPI ≤ target + significant volume
- Learner: Insufficient data or KPI close to target — continue
- Loser: Primary KPI > target × 1.5 with sufficient spend — kill
Phase 2: Deconstruct the Creative
Analyze the creative structure across 4 layers:
Hook (First second / First frame)
- What was used as the opener?
- Tactic identified (T01–T42)?
- Psychological trigger?
- Why did it work or not in light of the Hook Rate?
Body of the Message
- What creative mechanic (M1–M8)?
- What is the central promise?
- What proof point?
- Objection addressed or not?
CTA
- Clarity and alignment with the awareness stage?
- CTA friction adapted to engagement level?
Format & Production
- Format adapted to the platform and audience?
- Appropriate production quality (not too much, not too little)?
- Optimal ratio and duration?
Phase 3: Formulate the Diagnosis
Identify the primary cause of performance — good or bad.
If winner:
- What contributed most? (hook / angle / mechanic / format / audience)
- Is this result reproducible or contextual (period, budget)?
- What principle can be extracted for future creatives?
If loser:
- Where did the audience drop off? (Low Hook Rate = hook / Low CTR = body or CTA / High CPA = landing page or offer)
- Was the angle right but the execution poor? Or did the angle itself not work?
- What does this result reveal about the persona or market?
Phase 4: Generate 3 Test Hypotheses
Each hypothesis must be:
- Testable: one single variable changed
- Motivated: based on a signal in the data, not intuition alone
- Actionable: the next creative to produce is clear
Hypothesis format:
"If we [change X], then [KPI Y] should [increase/decrease] because [reason from the data]."
Output Format
# Ad Analysis — [Brand] — [Creative Name] — [Date]
## Performance Data
| Metric | Value | Benchmark | Status |
[table]
## Verdict
[ ] Winner [ ] Learner [ ] Loser
## Deconstruction
### Hook
Tactic: [T##] — Trigger: [bias/LF]
Analysis: [Why it worked / didn't work in light of the Hook Rate]
### Body of the Message
Mechanic: [M#] — Angle: [C#]
Analysis: [Reading of Hold Rate and CTR]
### CTA
Analysis: [Awareness alignment / clarity / friction]
### Format
Analysis: [Format, ratio, duration — adequacy]
## Primary Diagnosis
[1 sentence: what mainly explains the result]
## What We Learn
[2–3 generalizable insights for future creatives]
## 3 Test Hypotheses
**H1:**
If we [change X], then [KPI] should [direction] because [reason].
Creative to produce: [short description]
**H2:**
[Same structure]
**H3:**
[Same structure]
Related Skills
01-audit/brand-health — baseline before analysis
05-analysis/account-audit — macro view after multiple individual analyses
05-analysis/strategy-gap — if the analysis reveals untested angles
03-strategy/creative-brief — to brief the test hypotheses