Campaign Audit — Performance Marketing Agentic Workflow

schema_version: 1
version: "1.0.0"
id: campaign_audit
title: "Campaign Audit"
description: "Audit campaign structure, spend efficiency, tracking, creative, audiences, and bidding across all active channels. Returns prioritized findings with estimated impact."
overview: |
  A comprehensive audit chain that walks every active campaign across the account's paid channels and scores it across six dimensions: structure, spend efficiency, conversion tracking, creative quality, audience setup, and bid strategy. Each issue it surfaces is classified by severity and paired with an estimated performance impact if fixed.

  Use this at the start of a new engagement, before a quarterly review, or whenever performance has drifted without a clear cause. The output is a prioritised findings list designed to be read by a human media buyer, not acted on automatically.

category:
  level: task
  domain: marketing
  reusable: true
params:
  channels: string
  date_range: string
  media_context: string
steps:
  - type: llm.generate
    prompt: |
      You are a performance marketing analyst conducting a campaign audit.

      **Channels to audit:** {{channels}}
      **Date range:** {{date_range}}
      **Media context:** {{media_context}}

      Audit across these six dimensions:

      1. **Structure** — Campaign/ad group organization, naming conventions, segmentation quality
      2. **Spend Efficiency** — CPA/ROAS by segment, wasted spend identification, budget pacing
      3. **Tracking** — Conversion setup accuracy, platform vs backend discrepancy, UTM consistency
      4. **Creative** — Performance distribution, fatigue signals, testing cadence
      5. **Audience** — Targeting quality, overlap between ad sets, exclusion rules, funnel alignment
      6. **Bidding** — Strategy appropriateness, learning phase status, target realism

      For each dimension, score 0-100 using this rubric:
      - 90-100: Excellent — best practices followed, no issues
      - 70-89: Good — minor optimizations available
      - 50-69: Needs attention — clear waste or missed opportunities
      - 30-49: Poor — significant issues impacting performance
      - 0-29: Critical — fundamental problems, pause and fix before spending more

      Calculate the Overall Health Score as a weighted average:
      - Structure: 15% | Spend Efficiency: 25% | Tracking: 20% | Creative: 15% | Audience: 15% | Bidding: 10%

      Letter grade: A (90-100), B (80-89), C (70-79), D (50-69), F (<50)

      For each dimension provide:
      - Score (0-100) and letter grade
      - Finding: What specifically is wrong or at risk
      - Impact: Estimated waste or opportunity (in spend % or conversion lift %)
      - Action: Specific fix with priority (P1/P2/P3)

      **Quality Gate checks (auto-fail any dimension if violated):**
      - Tracking: If platform conversions differ from backend by >20%, score caps at 40
      - Spend: If >20% of spend goes to zero-conversion segments, score caps at 50
      - Bidding: If tCPA target is >50% away from actual CPA, score caps at 50

      Output format:
      ## Health Score: [0-100] ([Letter Grade])
      [One-line summary]

      ## Top 3 Findings
      [Prioritized by estimated impact]

      ## Dimension Scores
      [Table: Dimension | Score | Grade | Key Finding | Est. Impact]

      ## Prioritized Actions
      [Numbered list, P1 first, with estimated impact and effort level]

      ## Detailed Findings
      [Per-dimension breakdown with evidence]
    as: audit_report

  - type: llm.classify
    input: "{{audit_report}}"
    categories:
      - healthy
      - needs_optimization
      - urgent_issues
    prompt: "Based on the audit findings, classify the overall campaign health."
    as: health_status

  - type: variable.set
    name: result
    value: |
      **Campaign Audit Complete**
      Health: {{health_status}}

      {{audit_report}}