agentleFS
narayan-metaflownarayan-metaflow/metaflow-marketing-skills/skills/meta-ads-account-audit/SKILL.md

meta-ads-account-audit

Skill14 starsChanged 7 months agoMITView on GitHub

Systematic 75+ checkpoint audit based on the Sam Tomlinson framework: ICP validation → signal infrastructure → account structure → audiences → creatives → measurement — with ICE-prioritized recommendations.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Don't recommend drastic changes during learning phase: If an ad set launched <7 days ago, note it's in learning and defer judgment.

    line 225
  2. Distinctive:

    Flag Offline API migration: If the account references Offline Event Sets or Offline Conversions API, flag that it was discontinued May 2025 …

    line 229
  3. Distinctive:

    Confidence: If audit data is incomplete (no Events Manager access, missing creative-level frequency, limited date range), state what …

    line 233

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Process

  1. Collect data — Request or gather: Ads Manager export (90 days), Events Manager screenshots (Pixel + CAPI status, EMQ scores), creative performance breakdown, audience overlap reports, frequency data at creative level, and learning phase status per ad set
  2. Run the audit checklist — Work through each section below, flagging issues as Critical (red), Optimization (yellow), or Working Well (green)
  3. Score account health — Calculate overall health score based on findings
  4. ICE-prioritize recommendations — Score each finding: Impact (1-10) x Confidence (1-10) x Ease (1-10)
  5. Deliver action plan — Structured output with prioritized fixes and expected impact

Audit Checklist

Section 1: Signal Infrastructure

This is the foundation. Weak signals = weak delivery via Meta's Andromeda ranking system. Audit this first.

CheckpointWhat to CheckFlag If
Pixel activeMeta Pixel firing on all relevant pagesMissing or intermittent
CAPI activeConversions API implemented alongside PixelNot configured — this is always Critical in 2025
Event deduplicationMatching event_id between Pixel and CAPINot configured or double-counting visible
EMQ scoresEvent Match Quality per event typeBelow 6.0 (minimum acceptable). Target 8.0+
Matching keys sentEmail, phone, fbc, fbp, IP, user agentMissing high-impact keys (email, phone, fbc)
Conversion eventsCorrect events configured as optimization targetsOptimizing for micro-events (PageView) instead of macro-events (Purchase, Lead)
Event volumeConversions per ad set per weekBelow 25/week/ad set (stability minimum). Target 50+/week (optimal)
Conversion valuesDynamic values sent with Purchase eventsStatic or missing values for ecommerce
Offline trackingCRM/offline events flowing through CAPIUsing discontinued Offline Conversions API (killed May 2025) — migrate to standard CAPI with action_source: system_generated or physical_store
Attribution windows7-day click / 1-day view defaultNon-standard windows without clear rationale. Note: 7d_view and 28d_view removed from Insights API October 2025
AEM configurationAggregated Event Measurement for iOS 14+Event priority not configured, or low-value events prioritized over high-value

For detailed tracking fixes, use the meta-ads-tracking-setup skill.

Section 2: Account Structure

CheckpointWhat to CheckFlag If
Objective alignmentCampaign objectives match business goalsUsing Traffic objective when goal is Sales/Leads
Campaign consolidationBudget concentrated in fewer, well-funded campaignsToo many small-budget campaigns/ad sets (each needs 50+ conversions/week for optimal learning)
Campaign architectureClear structure: Prospecting, Retargeting, TestingNo separation between prospecting and retargeting
ASC usage (ecommerce)Advantage+ Shopping Campaigns for ecommerceEcommerce account with 100+ conversions/week not testing ASC (typically reduces cost per purchase ~12%)
New vs existing splitASC customer ratio configured intentionallyDefault ratio without analysis of actual customer mix
Naming conventionsConsistent, parseable taxonomy across campaigns/ad sets/adsInconsistent naming making analysis impossible
Special Ad CategoryHousing, employment, credit categories set where requiredMissing Special Ad Category when advertising in regulated industries

Section 3: Audience Strategy

CheckpointWhat to CheckFlag If
Audience size (TOF)Top-of-funnel audience reachBelow 200K for prospecting campaigns
Audience size (remarketing)Retargeting pool sizeBelow 1,000 (too small for stable delivery)
Audience overlapOverlap between ad sets in same campaignSignificant overlap causing self-competition
Exclusion strategyRecent converters and existing customers excluded from prospectingNo exclusions — wasting budget showing ads to existing customers
Targeting exclusions changeDetailed targeting exclusions removed March 31, 2025Still trying to use interest-based exclusions. Modern approach: customer list exclusions, recent converter exclusions, lifecycle suppression
Custom audiencesWebsite visitors, CRM uploads, engagement audiencesNo custom audiences built. CRM upload match rate benchmark: 40-70%
Lookalike seedsQuality of lookalike source audiencesLow-quality seeds (all website visitors) instead of high-quality seeds (purchasers, high-value customers). Start 1%, test up to 5-10%
Advantage+ audienceUsing algorithm-driven expansion with suggestionsNot testing Advantage+ audience on accounts with strong Pixel/CAPI data
Warm + cold mixing2025 best practice: combine in single ad setsHyper-segmented audiences in separate ad sets limiting algorithm learning
Engagement audiencesProactively building video viewers, page engagersNo engagement audiences — these only capture data from creation date, not retroactively

Section 4: Creative Health

CheckpointWhat to CheckFlag If
Creative frequencyFrequency at the creative level (not ad or ad set level)Frequency > 4 — Meta research: 45% conversion drop at 4 exposures
CTR trendCTR trajectory over 7-day rolling averageCTR declining > 20% from 7-day average = fatigue signal
CPC trendCPC movement correlated with CTRRising CPCs alongside declining CTR = confirmed fatigue
Creative volumeNumber of active creative variants per ad setFewer than 4 variants (insufficient for rotation) or more than 150 in ASC (hitting ceiling)
Format diversityMix of static, video, carousel, Reels-nativeSingle format only — missing performance from other formats
Aspect ratios4:5 vertical for Feed/Stories/Reels, 1:1 fallbackOnly landscape or only square creative
Video hooksFirst 3 seconds engagementNo bold hooks, silent-unfriendly (85% of Facebook videos watched without sound)
Ad copy lengthPrimary text effectivenessPrimary text exceeding 125 characters (gets truncated)
Refresh cadenceHow often new creative is introducedSame creatives running 30+ days for prospecting (target refresh every 7-14 days)
Landing page experiencePost-click experience qualitySlow load (>3 sec), message mismatch, poor mobile UX

For creative testing frameworks and fatigue management, use the meta-ads-creative-engine skill.

Section 5: Learning Phase Management

CheckpointWhat to CheckFlag If
Learning phase spend% of total account spend in learning phaseExceeds 20% of total spend. Meta: advertisers under 20% see up to 68% lower CPA
Ad set edits during learningSignificant changes within first 5-7 daysEdits resetting learning phase (budget changes >20%, audience changes, creative swaps, bid changes)
Event volume per ad setWeekly conversion volumeBelow 25/week — ad set may never exit learning
Ad set consolidationFewer, larger ad sets vs many small onesMultiple ad sets with <25 conversions/week each — consolidate
Learning limited statusAd sets stuck in "Learning Limited"Prolonged Learning Limited = restructure needed (consolidate audiences, increase budget, broaden optimization event)

Section 6: Budget & Bid Efficiency

CheckpointWhat to CheckFlag If
Budget allocationDistribution across campaigns and ad sets60/30/10 benchmark: 60% best performers, 30% scaling tests, 10% experiments
Bid strategyHighest Volume, Cost Cap, Bid Cap, ROAS GoalUsing Highest Volume without any guardrails when CPA/ROAS targets exist
Budget sufficiencyEnough budget for 50+ conversions/week/ad setBudget too low for optimization event (switch to higher-funnel event or increase budget)
PacingSpend trajectory vs budgetConsistently underspending (audience too narrow) or front-loading (audience too broad or bid too high)
Cost distributionSpend across ad sets within campaignsOne ad set consuming 90%+ of budget while others starve
Scaling methodHow budget increases are appliedIncreasing budget >20% at once (risks resetting learning)
CPM trendsCost per 1000 impressions trajectoryRising CPMs without corresponding performance improvement. Context: Meta avg ad price up 9% in 2025 (Q1 2025 earnings)

Section 7: Measurement & Attribution

CheckpointWhat to CheckFlag If
Attribution window7d click / 1d view defaultNon-standard without documented reason
API attribution change7d_view and 28d_view removed from Insights API October 2025Reporting still pulling deprecated windows
Cross-platform attributionConsistent measurement across Google + MetaDifferent windows making cross-channel comparison meaningless
Instant ExperienceNo longer counts as landing page view (June 2025)Funnel metrics inflated by counting Instant Experience as landing page views
Incrementality testingConversion lift or geo-based lift testsNo incrementality measurement — relying solely on platform-reported ROAS

Health Score Calculation

SectionWeightScore
Signal Infrastructure25%/10
Account Structure10%/10
Audience Strategy15%/10
Creative Health20%/10
Learning Phase10%/10
Budget & Bid Efficiency15%/10
Measurement & Attribution5%/10
Weighted Total/10

Interpretation: 8-10 = Healthy, 6-7.9 = Needs optimization, 4-5.9 = Significant issues, <4 = Critical.

Output Format

## Meta Ads Account Audit

### Health Score: [X.X / 10] — [Healthy / Needs Optimization / Significant Issues / Critical]

### Critical Issues (fix immediately)
1. **[Issue]** — [What's wrong, specific data] — ICE: [I×C×E = score]
   - Action: [Specific fix]
   - Expected impact: [Estimated improvement]

### Optimization Opportunities (fix within 2 weeks)
1. **[Issue]** — [What's wrong, specific data] — ICE: [score]
   - Action: [Specific fix]
   - Expected impact: [Estimated improvement]

### Working Well (maintain)
- [What's working and why]

### Priority Action Plan (next 30 days)
| Priority | Action | ICE Score | Timeline | Owner |
|----------|--------|-----------|----------|-------|
| 1 | [Highest ICE action] | [score] | Week 1 | [who] |

### 2025 Platform Change Flags
- [ ] CAPI implemented (mandatory in 2025)
- [ ] Offline Conversions API migrated (discontinued May 2025)
- [ ] Detailed targeting exclusions replaced with suppression strategy (removed March 2025)
- [ ] Instant Experience funnel metric adjustment (June 2025)
- [ ] 7d_view/28d_view API reporting updated (removed October 2025)
- [ ] ASC evaluated for ecommerce
- [ ] CPM trend accounted for (Meta avg +9% in 2025)

Example

Input: "I run a DTC skincare brand on Meta. Spending $8K/month across 6 campaigns. ROAS dropped from 4.2 to 2.1 over the last month. What's going on?"

Output:

Meta Ads Account Audit

Health Score: 4.8 / 10 — Significant Issues

Critical Issues

  1. Creative fatigue across all prospecting ad sets — Average creative frequency is 6.8 (threshold: >4). CTR dropped 34% over 30 days. Same 5 creatives running for 45+ days with no refresh. Meta research: 45% conversion drop at 4 exposures. — ICE: 10×9×7 = 630
    • Action: Introduce 4-6 new creative variants per ad set immediately. Mix formats (add video/Reels if currently static-only). Pause creatives with frequency >6
    • Expected impact: Meta data shows avg 8% conversion rate improvement when refreshing fatigued creatives. With this level of fatigue, impact likely higher
  2. No CAPI — Pixel only — Missing ~20-30% of conversion events. EMQ unavailable (no server events). Algorithm optimizing on incomplete data. — ICE: 9×9×8 = 648
    • Action: Implement CAPI (Shopify native if on Shopify, server-side GTM otherwise). See meta-ads-tracking-setup skill for full implementation guide
    • Expected impact: 20-30% more conversion signals → better optimization → lower CPA
  3. 6 campaigns with $8K/month = ~$44/day per campaign — Most ad sets getting <15 conversions/week. 4 of 6 campaigns stuck in Learning Limited. Meta: need 50+/week for optimal, 25+ minimum. — ICE: 9×8×6 = 432
    • Action: Consolidate to 3 campaigns max: (1) Prospecting, (2) Retargeting, (3) Testing. This concentrates budget for faster learning
    • Expected impact: Exiting learning phase → up to 68% lower CPA (Meta benchmark)

Optimization Opportunities

  1. No customer exclusions on prospecting — Existing customers seeing acquisition ads. Wasted spend + inflated ROAS (counting repeat purchases as new). — ICE: 7×8×9 = 504
    • Action: Upload customer list, exclude from prospecting campaigns. Detailed targeting exclusions removed March 2025 — use customer list suppression instead
    • Expected impact: 10-15% budget recovered from wasted impressions on existing customers
  2. All creatives are 1:1 square — Missing 4:5 vertical which outperforms 1:1 by ~15% in Feed (Billo data). No Reels-native content. — ICE: 6×7×6 = 252
    • Action: Create 4:5 versions of top performers. Test Reels-native vertical video
    • Expected impact: ~15% Feed performance improvement from aspect ratio alone
  3. No ASC campaign — DTC ecommerce brand with sufficient volume should test Advantage+ Shopping Campaigns. Typically reduces cost per purchase ~12% vs manual. — ICE: 7×6×5 = 210
    • Action: Launch ASC with top 10-20 creatives, set 70/30 new/existing split
    • Expected impact: ~12% CPA reduction if ASC outperforms manual (test with 30% of budget)

Working Well

  • Pixel is active and firing correctly on all pages
  • Attribution window set to 7d click / 1d view (correct default)
  • Product catalog connected and dynamic ads functional

Priority Action Plan (next 30 days)

PriorityActionICE ScoreTimelineOwner
1Implement CAPI648Week 1-2Dev + media buyer
2Refresh all creatives (new variants + formats)630Week 1-2Creative + media buyer
3Upload customer exclusion list504Day 1Media buyer
4Consolidate 6 campaigns → 3432Week 1Media buyer
5Create 4:5 vertical creative variants252Week 2-3Creative team
6Launch ASC test210Week 3-4Media buyer

2025 Platform Change Flags

  • CAPI implemented — NOT CONFIGURED (Critical)
  • Offline Conversions API — N/A (no offline events)
  • Detailed targeting exclusions — replaced with customer list suppression
  • Instant Experience — not using
  • 7d_view/28d_view — not pulling deprecated windows
  • ASC evaluated — not yet tested
  • CPM trend — Meta avg ad price up 9% in 2025, contributing to higher costs

Guidelines

  • Don't audit creative without frequency data at the creative level. Ad-level or ad-set-level frequency understates the problem. Meta's own research measures fatigue at the creative level across all ads/campaigns using that creative.
  • Don't recommend drastic changes during learning phase. If an ad set launched <7 days ago, note it's in learning and defer judgment. Changes during learning reset the clock.
  • Don't recommend hyper-segmented audiences. The 2025 Meta algorithm prefers consolidated audiences with conversion signals over manually sliced interest targeting. Combine warm + cold in single ad sets.
  • Don't ignore the 50/25 rule. 50 conversions/week/ad set is optimal; 25 is minimum for stability. If an ad set can't reach 25, recommend consolidating or optimizing for a higher-funnel event.
  • Account for CPM inflation. Meta's average ad price rose 9% in 2025. A ROAS decline may partly reflect higher costs, not just campaign degradation. Check CPMs alongside efficiency metrics.
  • Flag Offline API migration. If the account references Offline Event Sets or Offline Conversions API, flag that it was discontinued May 2025 and must migrate to standard CAPI.
  • Attribution API update. 7d_view and 28d_view windows removed from Insights API October 2025. Flag any reporting that still references these.
  • CTR benchmarks vary. WordStream September 2025 data shows Traffic campaigns average 1.71% CTR, but Lead and Sales campaigns vary significantly by industry. Say "benchmark varies by industry" rather than applying a single number.
  • Cross-references: For tracking fixes, use the meta-ads-tracking-setup skill. For creative testing frameworks, use the meta-ads-creative-engine skill. For Google Ads auditing, use the google-ads-audit skill. For cross-channel budget analysis, use the ad-spend-allocator skill.
  • Confidence: If audit data is incomplete (no Events Manager access, missing creative-level frequency, limited date range), state what conclusions are reliable and what needs more data.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.