Meta Ads Drop Diagnosis
診斷 Meta Ads 廣告系列中突然的成效下滑 — 找出 CPM、CTR、CPA 與曝光頻率的近期變化,以釐清根本原因
為什麼使用此範本
當您需要診斷 Meta Ads 廣告系列中突然的成效下滑時,使用此 diagnosis skill。
適用對象
希望執行 drop diagnosis 的 Meta Ads 廣告主與行銷人員。
使用方法
選擇此範本並填入您的廣告系列細節。AI 會逐步引導您完成 drop diagnosis 流程。
Required Credentials
| Credential | Where to Get | Used For | OAuth Scope |
|---|---|---|---|
| META_ACCESS_TOKEN | Meta Developer Console → Graph API Explorer → Generate Token | Fetching campaign performance data | ads_read (read-only) |
| META_AD_ACCOUNT_ID | Ads Manager URL: adsmanager.facebook.com/act_XXXXXXXXX | Identifying which account to query | — |
When to Use
Triggered by meta-ads-daily-pulse flagging a week-over-week drop > 20%, or when user reports "performance suddenly dropped." Not for gradual decline over weeks (that's creative fatigue territory).
Requirements
- account_id (required) — your META_AD_ACCOUNT_ID (format: act_XXXXXXXXX)
- campaign_id or ad_id of the declining entity (required) — find in Meta Ads Manager URL, or via GET /act_
{{AD_ACCOUNT_ID}}/campaigns?fields=id,name - date_range: last_7d (default)
Step 1: Establish the Drop
Confirm the drop with a Performance Snapshot mini-table:
| Metric | Last Week | This Week | WoW Change |
|---|---|---|---|
| Impressions | |||
| CPM | |||
| CTR | |||
| Frequency | |||
| CVR | |||
| CPA | |||
| ROAS |
Confirm it's real: WoW drop must be > 20% on at least 2 metrics to qualify as a "drop" (not normal fluctuation).
Step 2: Root Cause Analysis
Based on the metric pattern, identify the top 1-3 most likely root causes.
Use this decision tree:
| Pattern | Likely Root Cause |
|---|---|
| CTR down, frequency up | Creative fatigue |
| CTR down, frequency stable | Creative burnout or audience mismatch |
| CTR stable, CVR down | Landing page issue or offer problem |
| CTR up, CVR down | Landing page can't handle higher volume |
| Impressions down, CPM up | Competition increased, bid not competitive |
| Frequency up, ROAS down | Retargeting audience exhausted |
| All metrics down | External factor (seasonality, iOS, pixel issues) |
For each root cause, write 1-3 short sentences explaining what the data shows, why it points to this cause, and what other evidence supports or contradicts it.
Step 3: Recovery Plan
Format:
Immediate (next 24 hours):
- [Specific action: pause, scale down, refresh, shift budget]
- Why this helps now
Short-term (this week):
- [New creative, audience expansion, placement changes]
- What to test and why
Prevention (ongoing):
- [Rotation schedule, budget cap rules, monitoring thresholds]
- How to catch this earlier next time
Step 4: Output Format
Title: Summary Declining campaign: [Name] Performance Snapshot: [mini-table above]
Title: Root Cause Analysis [1-3 root causes with supporting evidence]
Title: Recovery Plan Immediate: [actions] Short-term: [actions] Prevention: [actions]
Rules
- Root cause must be supported by data — do not speculate without evidence
- If data is insufficient (e.g., < 3 days since change), say so and recommend waiting
- This is diagnosis only — specific action recommendations route to meta-ads-recommendation
- One immediate action maximum — don't overwhelm the user
Session Context — What This Skill Writes
After completing diagnosis, store the following in session context:
| Key | Description | Example |
|---|---|---|
| primary_root_cause | The most likely cause of the drop | "Creative fatigue — frequency 4.8, CTR dropped 38% WoW" |
| recovery_plan | Immediate + short-term + prevention actions | "Rotate new creative within 24h; set frequency cap at 3" |
meta-ads-recommendation reads these keys to produce the structured recovery action plan.