frank-luongt/skills/codex/campaign-analytics

<!-- AUTO-GENERATED by export-skills.py — DO NOT EDIT -->

---

name: campaign-analytics description: Measure multi-touch attribution, calculate channel ROI, analyze marketing funnels, and integrate A/B test results into campaign performance. Use when evaluating campaign effectiveness, optimizing channel spend, building attribution models, or reporting on marketing performance.

Campaign Analytics

Marketing campaign measurement framework — multi-touch attribution, channel ROI analysis, funnel diagnostics, and performance reporting. Turns campaign data into actionable spend allocation decisions.

Use this skill when

• Evaluating marketing campaign performance across channels
• Building or choosing a multi-touch attribution model
• Calculating ROI and ROAS per marketing channel
• Analyzing marketing funnel conversion at each stage
• Integrating A/B test results into campaign-level decisions
• Optimizing marketing budget allocation across channels
• Preparing marketing performance reports for leadership

Do not use this skill when

• Setting up analytics instrumentation from scratch (use analytics-tracking)
• Analyzing individual A/B test results (use ab-test-analysis)
• Optimizing a single landing page for conversion (use page-cro)
• Applying behavioral psychology to messaging (use marketing-psychology)
• Designing the campaign creative or strategy (use marketing agent skills)

Instructions

1. Select attribution model appropriate to your business (see model comparison).
2. Collect channel data — spend, impressions, clicks, conversions, revenue per channel.
3. Calculate channel ROI using the ROAS and incremental lift frameworks.
4. Analyze funnel conversion stage by stage to find drop-off points.
5. Integrate experiment results to quantify revenue impact of winning variants.
6. Produce the Campaign Performance Report with allocation recommendations.

---

Multi-Touch Attribution Models

Model Comparison

| Model | How Credit is Assigned | Best For | Limitation | |-------|----------------------|----------|------------| | First-Touch | 100% to first interaction | Understanding awareness drivers | Ignores nurture and conversion touches | | Last-Touch | 100% to last interaction before conversion | Understanding closing channels | Ignores awareness and nurture | | Linear | Equal credit to all touchpoints | Simple, unbiased baseline | No signal on which touches matter most | | Time-Decay | More credit to recent touchpoints | Long sales cycles with clear momentum | Undervalues awareness investments | | Position-Based (U-Shaped) | 40% first, 40% last, 20% middle | Balanced view of full funnel | Still arbitrary weight assignment | | Data-Driven (Algorithmic) | ML-based credit assignment | Large datasets, sophisticated teams | Requires significant data volume; black box |

Choosing the Right Model

Sales cycle < 7 days → Last-Touch (sufficient for short cycles)
Sales cycle 7-30 days → Position-Based (captures full journey)
Sales cycle > 30 days → Time-Decay or Data-Driven (long nurture matters)
Limited data (<1000 conversions/month) → Linear (unbiased baseline)
Abundant data (>5000 conversions/month) → Data-Driven (if tooling supports)

Attribution Implementation Checklist

• [ ] UTM parameters standardized across all channels
• [ ] Touchpoint tracking implemented (cookies, device graph, CRM matching)
• [ ] Attribution window defined (7-day click, 1-day view, 30-day click, etc.)
• [ ] Cross-device tracking configured (if applicable)
• [ ] Offline touchpoints included (events, sales calls, direct mail)
• [ ] Attribution model selected and documented
• [ ] Baseline established for comparison

---

Channel ROI Framework

Per-Channel Metrics Table

| Channel | Spend ($) | Impressions | Clicks | CTR | Conversions | CPA ($) | Revenue ($) | ROAS | |---------|-----------|-------------|--------|-----|-------------|---------|-------------|------| | Paid Search | | | | % | | | | X.Xx | | Paid Social | | | | % | | | | X.Xx | | Display / Programmatic | | | | % | | | | X.Xx | | Email | | | | % | | | | X.Xx | | Organic Search | | N/A | | N/A | | | | N/A | | Content / SEO | | N/A | | N/A | | | | N/A | | Referral / Partner | | | | % | | | | X.Xx | | Events / Webinars | | N/A | N/A | N/A | | | | X.Xx | | Total | $ | | | | | $ | $ | X.Xx |

Key Formulas

| Metric | Formula | Interpretation | |--------|---------|---------------| | ROAS | Revenue / Ad Spend | >3x = healthy for most B2B; >4x for e-commerce | | CPA | Total Spend / Conversions | Must be < Customer LTV for sustainability | | CAC | Total S&M Cost / New Customers | Include all costs (headcount, tools, agency fees) | | iROAS | Incremental Revenue / Incremental Spend | Measures true causal impact (not just correlation) | | Marginal ROI | Change in Revenue / Change in Spend | Use for budget reallocation — invest where marginal ROI is highest |

Incrementality Testing

Attribution models show correlation, not causation. Use incrementality tests to measure true impact:

| Method | How It Works | When to Use | |--------|-------------|-------------| | Geo-lift test | Run campaign in test geos, hold back control geos | Measuring offline + online impact | | Holdout test | Randomly exclude % of audience from ads | Measuring online display/social lift | | PSA (Ghost Ads) | Show public service ad instead of brand ad to control | Measuring brand lift without full holdout | | Pre/Post with control | Compare before/after with a control group | Quick directional read (less rigorous) |

---

Campaign Funnel Analysis

Funnel Stage Metrics

| Stage | Metric | Benchmark (B2B SaaS) | Benchmark (E-Commerce) | |-------|--------|---------------------|----------------------| | Awareness | Impressions, Reach, CPM | CPM: $5-15 | CPM: $2-8 | | Consideration | Clicks, CTR, CPC | CTR: 1-3%, CPC: $2-10 | CTR: 2-5%, CPC: $0.50-3 | | Conversion | Sign-ups, Purchases, CVR | CVR: 2-5% (trial), 1-3% (paid) | CVR: 1-4% | | Retention | Repeat rate, LTV | 30-day retention: 20-40% | Repeat purchase: 20-30% |

Funnel Diagnostic Template

## Campaign Funnel — [Campaign Name] — [Period]

| Stage | Volume | Rate | Benchmark | Gap | Diagnosis |
|-------|--------|------|-----------|-----|-----------|
| Impressions | | — | — | — | |
| Clicks | | CTR: % | % | | |
| Landing Page Views | | LPV rate: % | 90%+ | | |
| Conversions | | CVR: % | % | | |
| Revenue | $ | AOV: $ | $ | | |

### Biggest Drop-off
- **Stage:** [X] → [Y]
- **Expected rate:** [X]%
- **Actual rate:** [Y]%
- **Hypothesis:** [Why the drop-off occurred]
- **Recommended action:** [Specific intervention]

---

A/B Test Integration

Connecting Experiment Results to Campaign Decisions

When an A/B test completes, translate the result into campaign-level impact:

| Input | Source | Example | |-------|--------|---------| | Winning variant lift | A/B test analysis | +12% conversion rate | | Campaign conversion volume | Campaign data | 5,000 conversions / month | | Average conversion value | Revenue data | $50 per conversion |

Revenue Impact Calculation:

Incremental conversions = Current conversions x Lift %
                       = 5,000 x 0.12 = 600
Incremental revenue    = 600 x $50 = $30,000 / month
Annualized impact      = $30,000 x 12 = $360,000

When to Scale Winning Variants

| Confidence | Sample Size | Recommendation | |-----------|-------------|----------------| | >95% significance, >1000 conversions | Adequate | Scale to full traffic | | >90% significance, 500-1000 conversions | Borderline | Extend test 1 more week | | <90% significance | Insufficient | Do not scale — inconclusive |

---

Marketing Mix Modeling (MMM) — Overview

When to use MMM vs. Attribution:

| Dimension | Attribution | MMM | |-----------|-------------|-----| | Granularity | User-level | Channel-level aggregate | | Scope | Digital touchpoints | All channels (including offline, TV, OOH) | | Causation | Correlation-based | Regression-based (closer to causal) | | Latency | Real-time | Quarterly refresh | | Best for | Tactical optimization | Strategic budget allocation |

MMM is valuable when:

• Significant offline spend (events, TV, print, billboards)
• Need to model diminishing returns (saturation curves)
• Want to factor in seasonality and macroeconomic trends
• Budget allocation decisions across 5+ channels

---

Output Template: Campaign Performance Report

# Campaign Performance Report — [Period]

## Executive Summary
- Total spend: $[X] across [Y] channels
- Total revenue attributed: $[X]
- Blended ROAS: [X]x
- Key insight: [1 sentence]

## Channel Performance
[Per-Channel Metrics Table from above]

## Top Performing Campaigns
| Campaign | Channel | Spend | Revenue | ROAS | Key Driver |
|----------|---------|-------|---------|------|------------|
| | | $ | $ | X.Xx | |

## Funnel Analysis
[Funnel diagnostic with biggest drop-off identified]

## Attribution Insights
- Model used: [model name]
- Top converting paths: [e.g., Paid Search → Email → Direct]
- Undervalued channels: [channels receiving less credit than expected]

## Budget Recommendation
| Channel | Current Spend | Recommended Spend | Change | Rationale |
|---------|--------------|-------------------|--------|-----------|
| | $ | $ | +/-% | |

## Next Period Plan
1. [Action] — Expected impact — Owner
2. [Action] — Expected impact — Owner

---

Common Mistakes

• Attributing 100% credit to last touch — over-invests in bottom-funnel at the expense of awareness
• Comparing channels without controlling for intent — branded search has high conversion because of pre-existing intent, not because the ad is effective
• Ignoring incrementality — correlation is not causation; run holdout tests before making large budget shifts
• Reporting vanity metrics — impressions and clicks don't pay the bills; report on revenue, ROAS, and CPA
• Optimizing for CPA alone — the cheapest leads are often the lowest quality; optimize for CAC payback or LTV:CAC
• No attribution window discipline — without a defined window (7-day, 30-day), you'll double-count conversions

---

Additional Resources

• Related skills: ab-test-analysis (experiment-level analysis), analytics-tracking (instrumentation), page-cro (landing page optimization), marketing-psychology (behavioral science for messaging)
• Google's Marketing Mix Model (Meridian) — open-source MMM framework
• Meta's Robyn — open-source MMM library

<!-- Source: .faos/custom/skills/business/campaign-analytics/SKILL.md -->