Good Marketing Skill

Digital Marketing Analytics Guide

Overview

This skill guides the agent through proper digital marketing measurement, attribution modeling, customer analytics, and campaign optimization. It covers the full customer journey from awareness through retention, ensuring methodologically sound measurement and privacy-compliant data practices.

When to Use

Use this skill when the user wants to:

• Set up or evaluate marketing attribution models
• Calculate and predict customer lifetime value (CLV/LTV)
• Measure campaign ROI/ROAS with incrementality testing
• Optimize SEO/SEM strategy with proper keyword research
• Design email marketing campaigns with deliverability best practices

Decision Tree

IF measuring marketing channel effectiveness:
  IF the goal is real-time campaign optimization:
    → Use multi-touch attribution (MTA) with model awareness
    → WARN: No single attribution model is "correct" — always report
      multiple models (last-click, linear, time-decay, data-driven)
    → Acknowledge model bias: last-click overvalues bottom-funnel,
      first-click overvalues top-funnel
  ELIF the goal is strategic budget allocation:
    → Use Marketing Mix Modeling (MMM) with adstock/carryover effects
    → Model diminishing returns via Hill functions or log transforms
    → Account for cross-channel synergy and interaction effects
  ELIF the goal is causal impact measurement:
    → Use incrementality testing (geo experiments, holdout groups)
    → This is the gold standard for establishing causal lift

IF calculating customer value:
  IF contractual business (subscription, SaaS):
    → Use survival models with discount rate and time horizon
    → Define churn operationally (cancellation, non-renewal)
  ELIF non-contractual business (e-commerce, retail):
    → Use probabilistic models (BG/NBD for frequency, Gamma-Gamma
      for monetary value)
    → Define inactivity window threshold with justification
    → Account for recency, frequency, and monetary value (RFM)

IF optimizing search:
  IF organic search (SEO):
    → Ensure technical SEO fundamentals (crawlability, indexation,
      Core Web Vitals — LCP, CLS, INP)
    → Address E-E-A-T (Experience, Expertise, Authoritativeness,
      Trustworthiness) and topical authority
    → Keyword research: balance search volume with user intent
      (informational, transactional, navigational) and difficulty
  ELIF paid search (SEM/PPC):
    → Evaluate bidding strategies: manual CPC vs. tCPA vs. tROAS
    → Consider quality score optimization alongside bid management
    → Track cost per click (CPC) and ROAS as primary efficiency metrics

Guardrails

• REFUSE to recommend last-click attribution as the correct or only model.

Always present model assumptions and limitations.

• REFUSE to calculate CLV as simply "average revenue × average lifespan"

without specifying a probabilistic or contractual model framework.

• WARN if marketing KPIs are vanity metrics (followers, likes, impressions)

not tied to business outcomes (revenue, profit, LTV). Marketing measurement should ladder up to north star metrics with clear alignment to business OKRs.

• WARN if tracking is discussed without addressing privacy compliance

(GDPR, CCPA, iOS ATT, cookieless measurement). First-party data strategy and consent management are non-negotiable.

• MUST NOT recommend purchasing email lists. Always require opt-in consent

and proper list hygiene (bounce removal, suppression lists, re-engagement campaigns for inactive subscribers).

• MUST NOT ignore deliverability — SPF, DKIM, and DMARC authentication

are table stakes for email marketing.

Attribution Model Comparison

| Model | Bias | Best For | |-------|------|----------| | Last-click | Overvalues bottom-funnel (branded search) | Quick directional signal | | First-click | Overvalues top-funnel (display, social) | Awareness campaign eval | | Linear | Dilutes across all touchpoints | Even-handed but naive | | Time-decay | Overweights recent touches | Conversion-focused | | Data-driven | Requires significant data volume | Best available observational model | | Incrementality | Gold standard — causal | Strategic decisions |

Privacy & Compliance

All digital marketing measurement must account for:

• GDPR (EU) and CCPA/CPRA (California) consent requirements
• iOS 14+ ATT framework limiting IDFA availability
• The shift from third-party cookies to first-party data strategies
• Server-side tracking (Conversion API / CAPI) as complement to pixel-based
• Cookie consent banners and their impact on data completeness
• Privacy Sandbox proposals (Topics API, Attribution Reporting API)

Tracking Infrastructure

• Use UTM parameters consistently for campaign attribution
• Implement tag management (Google Tag Manager) for governance
• Consider server-side tracking for reliability and privacy compliance
• Set up proper conversion tracking with defined conversion windows
• Use event-based measurement (GA4 paradigm) over pageview-based

Ad Tech & Programmatic

When working with programmatic advertising:

• Understand auction mechanics: first-price auctions require bid shading
• Set frequency caps to prevent ad fatigue and wasted spend
• Monitor viewability (MRC standard: 50% pixels, 1+ second) and

invalid traffic (IVT) to ensure ad verification

• Evaluate brand safety controls alongside performance metrics

Email Marketing Best Practices

• Prioritize deliverability: authenticate with SPF, DKIM, DMARC
• Practice list hygiene: remove hard bounces, re-engage or sunset

inactive subscribers, honor unsubscribes promptly

• Personalization should use behavioral signals (purchase history,

browsing behavior, engagement patterns, lifecycle stage) — not just dynamic name insertion

• Segment by engagement level, lifecycle stage, and purchase behavior

Content & SEO Strategy

• Build topical authority through pillar-cluster content architecture
• Focus on E-E-A-T signals: demonstrate real experience, expertise,

authoritativeness, and trustworthiness

• Prioritize user experience and helpful content over keyword density
• Address Core Web Vitals: LCP, CLS, INP for page experience

Customer Analytics

• Segmentation: Use data-driven methods (RFM scoring, k-means clustering

with silhouette validation) rather than arbitrary thresholds

• Cohort analysis: Normalize by initial cohort size and control for

seasonality effects across time periods

• Churn modeling: Define churn criteria operationally; use recency-based

thresholds for non-contractual settings

Edge Cases

• Small advertisers: May not have enough conversion volume for

data-driven attribution — fall back to rules-based with caveats

• Cross-device journeys: Acknowledge deterministic matching gaps;

probabilistic identity graphs are incomplete

• Offline conversions: Store visits, phone calls require modeling

or direct integration (store visit conversions, call tracking)

• View-through attribution: High inflation risk in cookieless

environments — require incrementality validation

Escape Hatch

Experienced marketers can override guardrails by explicitly stating: "I acknowledge [specific guardrail] and am proceeding because [justification]."

References

For detailed methodology guides, see the references/ directory.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: WilliamWJHuang
• Source: WilliamWJHuang/ab-test-causal-inference-skills
• License: MIT

Install and usage instructions live in the source repository linked above.