The CFO Question That Changed Everything
Years ago, I presented my annual marketing spend to the CFO. She asked a simple question: “Which channels actually drove revenue?” I pulled up the data—Google Ads showed $4 million in attributed sales, Facebook showed $2.3 million, email showed $800,000, and affiliates showed $1.2 million. Her follow-up: “So which should we increase?” The answer should have been obvious.
But then I actually thought about the customer journey. A typical customer might first see a Facebook ad, click through to browse, leave, return through a Google search for our brand, browse more, get an email abandonment sequence, wait a week, click through, add to cart, leave again, then finally return via a retargeting ad and purchase. Which channel gets credit? Every channel touched that customer. None of them alone created the purchase.
That’s when I understood: attribution is the most important question in marketing, and no one model is perfect.
Why Attribution Matters
Attribution determines which marketing activities get credit for results. That credit determines budget allocation. Budget allocation determines growth. Get attribution wrong, and you defund winners while funding losers—eventually destroying your competitive position.
The average customer journey involves 8-15 touchpoints before purchase. With multiple channels, devices, and time passing between those touchpoints, correctly assigning credit is both critically important and genuinely difficult.
The brands winning at marketing aren’t necessarily spending the most—they’re spending smarter. Attribution provides the intelligence for that smarter spending.
The Attribution Model Spectrum
Attribution models exist on a spectrum from simple to complex:
Single-Touch Models (Simple)
- First-touch: Credit to first interaction
- Last-touch: Credit to final interaction
Multi-Touch Models (Complex)
- Linear: Equal credit across all touchpoints
- Time-decay: More credit to recent touches
- Position-based: Credit at ends, middle gets less
- Algorithmic: Data-driven model distribution
Understanding Each Model
First-Touch Attribution
100% credit goes to the first channel a customer interacts with. Best for understanding awareness creation.
- Strengths: Simple to track, identifies awareness channels, clear decision
- Weaknesses: Ignores nurturing, overvalues top-of-funnel, punishes retargeting
- Best for: Brand awareness campaigns, new product launches
Last-Touch Attribution
100% credit goes to the final channel before purchase. Best for understanding conversion optimization.
- Strengths: Identifies direct response channels, clear conversion optimization path
- Weaknesses: Ignores awareness and nurture, undervalues research phases
- Best for: Direct response campaigns, bottom-funnel optimization
Linear Attribution
Equal credit across every touchpoint in the journey. Respects every channel’s contribution.
- Strengths: Fair distribution, acknowledges all touchpoints
- Weaknesses: Undervalues high-intent touches, doesn’t reflect actual influence
- Best for: Balanced reporting, customer education products
Time-Decay Attribution
More credit to touches closer in time to purchase. Model reflects that recent touches drove the decision.
- Strengths: Reflects reality of purchase decisions, practical optimization
- Weaknesses: Still arbitrary weighting, may undervalue awareness
- Best for: Considerate purchase journeys, B2B sales
Position-Based (U-Shaped) Attribution
40% credit each to first and last touch, remaining 20% distributed across middle touches.
- Strengths: Balances awareness and conversion, acknowledges first/last importance
- Weaknesses: Arbitrary percentages, may not fit all journeys
- Best for: Most B2C e-commerce
Algorithmic Attribution
Data-driven model using machine learning to determine credit distribution based on actual conversion patterns.
- Strengths: Customized to actual data, more accurate
- Weaknesses: Requires data volume, requires technical setup
- Best for: Large data volumes, sophisticated teams
Pro Tip: Use Multiple Models
Ask different questions, get different answers. Run your attribution using at least three models: first-touch (what creates awareness), last-touch (what drives conversion), and algorithmic (what actually works). Compare and contrast. The differences reveal insights no single model can provide. No model is “correct”—they’re all approximations of an unknowable reality.
Common Mistakes to Avoid
- Using only last-touch: Destroying top-of-funnel budgets by undervaluing awareness
- Trusting one model completely: No model perfectly represents customer reality
- Ignoring cross-device tracking: Missing journey touchpoints across devices
- Not cleaning data: GIGO applies: garbage attribution data = garbage insights
- Changing models frequently: No model comparisonability over time
- Not tracking offline: Ignoring in-store, phone, and person-to-person touchpoints
- Overcomplicating: Need enough data to support whatever model you choose
Implementation Requirements
1. Tracking Infrastructure
- UTM parameter implementation
- Cross-device tracking capabilities
- CRM integration
- Offline conversion tracking
- Customer journey visualization
2. Data Quality Requirements
- Clean data with proper categorization
- Consistent naming conventions
- Accurate conversion tracking
- Proper session tracking
- Sufficient data volume
3. Platform Considerations
Most marketing platforms have attribution built in:
- Google Analytics: Multiple models, cross-platform tracking
- Meta Ads: Conversion-based, view-through, click-through
- LinkedIn: Account-based, engagement-based
- Shopify: Built-in attribution for merchants
Dedicated solutions: Rockerbox, Northbeam, AppsFlyer for mobile.
FAQ: Attribution Modeling
Q: Which attribution model is best?
A: There is no perfect model. Different models answer different questions. First-touch answers awareness creation. Last-touch answers conversion. Algorithmic answers actual performance. Use all of them.
Q: How much data do we need for algorithmic attribution?
A: Algorithmic models need thousands of conversions to produce reliable results. Without sufficient data, choose simpler models. Better conservative (first-touch) than wrong (algorithmic with poor data).
Q: Should we track across devices?
A: Yes, absolutely—if your customer journey crosses devices. For most e-commerce, at least 20-30% of journeys involve multiple devices. Without cross-device tracking, you’re missing significant touchpoints.
Q: How often should we review attribution?
A: Monthly for trend monitoring, quarterly for model reviews. Significant changes in customer behavior or marketing strategy warrant model review.
Q: Can we attribute offline sales?
A: Yes, with proper tracking: unique promo codes, survey questions (“how did you find us?”), CRM matching, and store attribution links.
Attribution Modeling Checklist
- Define conversion events to track
- Implement proper UTM tracking
- Set up cross-device tracking
- Clean and categorize data
- Choose initial models (at least 2)
- Establish baseline attribution
- Track by device and channel
- Monitor customer journey paths
- Review model performance quarterly
- Document learnings and insights
- Align attribution with business goals
Final Thoughts
Attribution isn’t a destination—it’s a journey. Your understanding of what drives results will evolve as your data grows and your customer journeys become clearer. Start with simple models, build your tracking infrastructure, and graduate to more sophisticated approaches as your program matures.
Remember: every model is wrong in some way. The question is which wrong model provides the most useful approximation for your business decisions. First-touch tells you about awareness. Last-touch tells you about conversion. Everything in between tells you the full story. Don’t let perfection become the enemy of useful progress.
The models don’t matter as much as consistent analysis over time. Pick models that help you make better decisions, apply them consistently, and evolve as your data and capabilities grow.
