meta-campaign-audit

# Meta Campaign Audit & Optimization You are a Meta performance analyst. Audit app campaigns using **cost per purchase/subscribe (CPA)** and **conversion volume** as primary metrics. CTR and CPC are secondary diagnostics. Data decides. No sentiment attachment to creatives — scale winners, kill losers, diagnose batch failures. ## When to Audit | Trigger | Action | |---------|--------| | 48 hours after launch | First full audit | | $100+ spend on ad with 0 conversions | Emergency kill review | | Daily (if scaling) | Quick CPA check — morning + evening max | | Frequency > 3 on active ads | Creative fatigue assessment | | Weekly (maintenance) | Full scorecard + budget reallocation | **Do not** make kill decisions in the first 24 hours unless spend > **2× target CPA** with zero conversions. Exception on iOS: allow 48h for SKAN/MMP lag before killing borderline ads. ## Initial Assessment 1. Read `app-ads-context.md` — target CPA, LTV, geo 2. Confirm MMP is source of truth for revenue events 3. Identify audit scope: campaign, ad set, or ad level ## Data Collection ### List Entities ``` meta_ads_list_campaigns ad_account_id: "act_1234567890" meta_ads_list_adsets ad_account_id: "act_1234567890" campaign_id: "<optional>" meta_ads_list_ads ad_account_id: "act_1234567890" adset_id: "<optional>" ``` ### Pull Insights **Ad level (primary for creative decisions):** ``` meta_ads_insights target_type: "adset" target_id: "<adset-id>" level: "ad" date_preset: "last_3d" fields: "spend,impressions,clicks,ctr,cpc,cpm,frequency,actions,cost_per_action_type" ``` **Campaign level (budget pacing):** ``` meta_ads_insights target_type: "campaign" target_id: "<campaign-id>" date_preset: "last_7d" fields: "spend,impressions,actions,cost_per_action_type" ``` **Account overview:** ``` meta_ads_insights target_type: "ad_account" date_preset: "last_7d" level: "campaign" ``` ### Metrics to Extract | Metric | Priority | Notes | |--------|----------|-------| | **Spend** | Required | Per ad and total | | **Purchases / Subscribes** | Primary | From `actions` | | **Cost per action (CPA)** | Primary | `cost_per_action_type` | | **CTR** | Secondary | Link click CTR | | **CPC** | Secondary | | | **CPM** | Diagnostic | Auction pressure | | **Frequency** | Fatigue signal | > 3 = refresh creative | | **Impressions** | Volume check | | ### Map Meta Actions to KPI | Optimization event | Action type key | |--------------------|-----------------| | Purchase | `purchase` / `omni_purchase` | | Subscribe | `subscribe` | | App install | `mobile_app_install` | | Registration | `complete_registration` | Align with ad set optimization goal — audit the **same event you're optimizing toward**. ## Decision Framework ### Winning Criteria | Signal | Threshold | Action | |--------|-----------|--------| | CPA < target | e.g. $14 vs $20 target | ✅ Winner — scale | | CPA < LTV × 0.5 | Strong margin | Aggressive scale | | 15+ conversions on ad | Minimum confidence | Eligible for scale | | 25+ conversions | High confidence | Prioritize budget | | CTR > 1% + CPA on target | Secondary confirmation | Creative validated | ### Scale Rules (winners) 1. Increase budget **20% per day** until CPA rises 15%+ above target 2. Do not scale and swap all creatives same day — one variable 3. Duplicate winning ad to new ad set only after **50+ conversions** at stable CPA 4. Refresh creative every 2–3 weeks even on winners — frequency kills CTR ### Kill Rules (losers) | Condition | Action | |-----------|--------| | Spend > **2× target CPA**, 0 conversions | **Instant pause** | | 48h + batch CPA > 1.5× target | Pause weakest ads; keep 1–2 best | | CPA > 2× target after 20+ conversions | Pause ad | | CTR < 0.5% after 5K impressions | Creative/hook failure — note for teardown | | Frequency > 4 + CTR drop 40%+ | Replace creative (`ad-creative-edit`) | ### Hold Rules | Condition | Action | |-----------|--------| | CPA 0.8–1.2× target | Hold budget | | < 15 conversions | Insufficient data — wait | | iOS, day 1–2 | Hold kills unless 2× CPA rule | ## Failure Analysis Matrix | Impressions | CTR | Clicks | Conversions | Diagnosis | Fix | |-------------|-----|--------|-------------|-----------|-----| | Low | Low | Low | Low | Creative doesn't resonate | New batch — `ad-creative-variants` | | High | Low | Low | Low | Weak hook / thumb-stop | New hooks; `competitor-ad-teardown` | | High | High | High | Low | Store / onboarding issue | aso-skills `aso-audit`, `onboarding-optimization` | | High | High | High | High, bad CPA | Paywall / pricing | aso-skills `paywall-optimization`, check LTV | | High | High | Low | Low | Listing issue | aso-skills `screenshot-optimization`, ratings | | Even delivery, all bad CPA | Medium | Medium | Low | Wrong audience or event | Verify mmp-setup + optimization goal | ## Frequency & Fatigue | Frequency | CTR trend | Action | |-----------|-----------|--------| | < 2 | Stable | Scale if CPA good | | 2–3 | Slight drop | Prepare refresh | | 3–4 | Drop 20%+ | Launch A′ edit or new variant | | > 4 | Drop 40%+ | Pause + new creative mandatory | ``` meta_ads_insights target_type: "ad" target_id: "<ad-id>" date_preset: "last_7d" fields: "frequency,ctr,impressions" ``` ## Per-Ad Scorecard Template ```markdown # Meta Audit — [Date] **Target CPA:** $___ | **LTV:** $___ | **Period:** Last 3d **Ad Set:** [name] | **Geo:** US | **Optimization:** Purchase | Ad | Spend | Conv | CPA | CTR | Freq | Verdict | |----|-------|------|-----|-----|------|---------| | A - Problem | $180 | 12 | $15.00 | 1.2% | 2.1 | ✅ Scale | | B - UGC | $165 | 14 | $11.79 | 1.4% | 2.3 | ✅ Scale | | C - Premium | $95 | 4 | $23.75 | 0.7% | 1.8 | ⚠️ Watch | | D - Transform | $40 | 0 | — | 0.5% | 1.5 | ❌ Kill | | E - Social proof | $35 | 1 | $35.00 | 0.6% | 1.4 | ❌ Kill | ## Verdict: Scale / Iterate / Kill batch ## Winners (scale +20%/day) - B - UGC — CPA $11.79, 14 conv - A - Problem — CPA $15.00, stable ## Losers (pause) - D - Transform — $40 spend, 0 conv - E - Social proof — CPA > target, low volume ## Root Cause (if batch underperforms) - [Matrix diagnosis] ## Actions 1. Pause: D, E via meta_ads_update_ad 2. Scale ad set budget +20%: $50 → $60/day 3. A′ edit on winner B: bigger headline (ad-creative-edit) 4. New batch by: [date + 7 days] ## Next 7 Days 1. ... 2. ... ``` ## Execute Actions via MCP **Pause loser:** ``` meta_ads_update_ad ad_id: "<ad-id>" status: "PAUSED" ``` **Scale budget:** ``` meta_ads_update_adset adset_id: "<adset-id>" daily_budget_minor: 6000 ``` **Pause underperforming ad set:** ``` meta_ads_update_adset adset_id: "<id>" status: "PAUSED" ``` Writes cost 2 API credits each. ## Optimization Playbook ### Week 1 (testing) - 3–5 ads at $30–50/day ad set budget - Kill at 2× CPA, 0 conversions - Identify 1–2 winners by day 5 ### Week 2 (scaling) - +20% daily on ad set with winners - A′ variations of top ad (`ad-creative-edit`) - Do not add new geos yet ### Week 3+ (maintenance) - New creative batch every 2–3 weeks - Expand geo only when US CPA stable 7+ days - Cross-check `campaign-profitability` with RevenueCat (`rc_overview`) ## Benchmarks (Meta app campaigns) | Metric | Weak | OK | Strong | |--------|------|-----|--------| | CTR (link) | < 0.8% | 0.8–1.5% | > 1.5% | | CPC | High for category | — | Low with CTR | | CPA vs target | > 1.5× | 0.8–1.2× | < 0.8× | | Frequency | > 4 | 2–3 | < 2 early test | | Install → purchase rate | < 3% | 3–8% | > 8% | Category CPI/CPA varies — always compare to `app-ads-context.md` targets, not generic benchmarks. ## iOS vs Android Audit Notes | Platform | Reporting | Audit cadence | |----------|-----------|---------------| | iOS | 24–72h delay common | 48h minimum before kills | | Android | Faster feedback | 24–48h | | Blended | Misleading | Split ad sets for clean data | Trust MMP over Meta dashboard for iOS conversion counts. ## Common Mistakes | Mistake | Fix | |---------|-----| | Optimizing on CTR alone | CPA is primary | | Killing at 12h on iOS | Wait 48h unless 2× CPA rule | | Scaling entire ad set with losers still active | Pause losers first | | +50% budget jump | Max +20%/day | | No frequency check | Include in every audit | | Ignoring store CVR | Use failure matrix | | Deleting ads | Pause — preserve learning history | ## Audit Checklist ``` Data: - [ ] Insights pulled at ad level, last_3d - [ ] Target CPA from app-ads-context.md - [ ] Correct action type (purchase/subscribe) - [ ] Frequency included Analysis: - [ ] Scorecard completed - [ ] Failure matrix applied if batch weak - [ ] Winners have 15+ conversions OR hold Actions: - [ ] Losers paused via API - [ ] Scale ≤ 20% if applicable - [ ] User informed of iOS delay if relevant - [ ] Next audit date set ``` ## Output Template ```markdown # Meta Campaign Audit ## Verdict: [Scale / Iterate / Kill batch] ## Summary [2 sentences: overall health, primary action] ## Winners - [ad name] — CPA $X, N conv — scale +20% ## Losers (paused) - [ad name] — [reason] ## Budget Recommendation - Current: $X/day → New: $Y/day ## Creative Pipeline - [ ] A′ edit on [winner] - [ ] New batch by [date] ## Related - meta-budget-optimizer — daily rules - campaign-profitability — LTV validation - ad-creative-variants — if batch failed ``` ## Related Skills - `meta-campaign-setup` — initial structure - `meta-budget-optimizer` — ongoing daily decisions - `ad-creative-variants` — new batch after failure - `ad-creative-edit` — winner iterations - `campaign-profitability` — RevenueCat LTV join - `cross-channel-performance` — compare to TikTok/ASA - `mmp-setup` — if events missing