Linear vs. Non-Linear Attribution Models for Paid Acquisition

The Attribution Problem in Paid Acquisition

Attribution modeling is not a reporting nicety—it's the foundation of paid acquisition economics. Every dollar you allocate to TikTok, Meta, or Reddit depends on understanding which touchpoint actually drove the conversion. Get this wrong, and you're optimizing toward phantom signals while starving high-ROI channels of budget.

The core tension: customers interact with your brand 20–500 times before converting[1]. A single-touch model (first-click or last-click) assigns 100% credit to one interaction, ignoring the other 19–499. Linear attribution attempts to solve this by distributing equal credit across all touchpoints[1][2]. But equal distribution assumes all interactions have equal impact—a mathematically convenient fiction that destroys profitability at scale.

This post dissects the economic reality of linear vs. non-linear models, with specific frameworks for paid acquisition optimization.

---

Core Economic Principles: Why Attribution Matters

Attribution directly impacts three critical metrics:

Customer Acquisition Cost (CAC): [ \text{CAC} = \frac{\text{Total Marketing Spend}}{\text{Conversions Attributed}} ]

Return on Ad Spend (ROAS): [ \text{ROAS} = \frac{\text{Revenue from Attributed Conversions}}{\text{Ad Spend}} ]

Lifetime Value to CAC Ratio: [ \text{LTV:CAC} = \frac{\text{Customer Lifetime Value}}{\text{CAC}} ]

Under linear attribution, if a customer touches five channels before converting, each channel receives 20% credit. This creates a critical distortion: channels that appear early in the funnel (awareness) receive inflated credit, while channels that close deals (conversion) receive deflated credit.

Example: A customer sees a Facebook ad (awareness), engages with email (nurture), clicks a Google search ad (intent), and converts. Linear attribution gives each 33% credit. But the Google search ad likely drove 70%+ of conversion probability. By crediting Facebook equally, you're systematically underfunding high-intent channels and overfunding low-intent awareness.

The economic consequence: misallocated budget reduces overall ROAS and increases CAC.

---

Linear Attribution: The Oversimplified Model

Definition: Linear attribution distributes 100% of conversion credit equally across all touchpoints[1][2][3].

Mathematical Structure: [ \text{Credit per Touchpoint} = \frac{100%}{\text{Number of Touchpoints}} ]

If a customer has 5 touchpoints: each receives 20% credit.

When Linear Attribution Fails

1. Over-crediting Low-Impact Interactions

Not all touches are equal[4]. A user opening a promotional email out of habit receives the same credit as a high-intent webinar attendance. Linear attribution treats passive interactions (accidental clicks, routine opens) identically to active engagement (form submissions, demo requests)[4].

2. Ignoring Touchpoint Strength

A well-targeted webinar, case study, or sales call has exponentially higher conversion influence than a display impression[4]. Linear attribution assigns equal weight, failing to recognize which interactions actually move buying decisions[4].

3. Unsuitable for Long Sales Cycles

In B2B with 6–12 month sales cycles involving content downloads, demos, multiple calls, and nurturing campaigns, linear attribution provides only a "broad-strokes view"[1]. It acknowledges each touchpoint exists but obscures which interactions were truly pivotal[4].

4. Dilutes Conversion Credit

By spreading credit equally, linear attribution makes it difficult to identify which channels genuinely drive conversions[1]. A channel that appears in 50% of customer journeys but drives only 10% of conversions receives inflated credit.

Accessibility Trade-off

Linear attribution's primary advantage: it's accessible[4]. It requires no custom models, deep integrations, or data science resources. Most analytics platforms include it out-of-the-box[4]. For teams without infrastructure to track weighted contributions, linear is simpler than nothing[4].

But simplicity is not a business advantage—accuracy is.

---

Non-Linear Attribution Models: Precision-Based Allocation

Non-linear models weight touchpoints based on actual conversion influence. The primary categories:

1. Time-Decay Attribution

Principle: Touchpoints nearest conversion receive highest weight[1].

Mathematical Structure: [ \text{Credit} = \frac{\text{Recency Weight}}{\text{Sum of All Weights}} ]

Example with exponential decay (half-life = 7 days):

• Touchpoint 1 (30 days ago): weight = 0.5^(30/7) ≈ 0.016
• Touchpoint 2 (14 days ago): weight = 0.5^(14/7) ≈ 0.25
• Touchpoint 3 (7 days ago): weight = 0.5^(7/7) ≈ 0.5
• Touchpoint 4 (1 day ago): weight = 0.5^(1/7) ≈ 0.91

Total weight = 1.726. Normalized credits: 0.9%, 14.5%, 29%, 52.7%

Use Case: Short-cycle conversions (e-commerce, SaaS free trials). The closer a touchpoint to purchase, the more influence it had.

Limitation: Underweights awareness-stage activities that build brand consideration. For 90-day sales cycles, this model may miss critical early-funnel influence.

2. Position-Based (U-Shaped) Attribution

Principle: First and last touchpoints receive 40% credit each; middle touchpoints split remaining 20%[1][5].

Mathematical Structure: [ \text{First Touch Credit} = 40% ] [ \text{Last Touch Credit} = 40% ] [ \text{Middle Touches Credit} = \frac{20%}{\text{Number of Middle Touchpoints}} ]

Example with 5 touchpoints:

• First: 40%
• Middle 1: 6.67%
• Middle 2: 6.67%
• Middle 3: 6.67%
• Last: 40%

Use Case: B2B demand generation where awareness (first touch) and conversion (last touch) are distinct stages. Middle touches (nurture) support progression but don't drive conversion alone.

Limitation: Arbitrary 40/40/20 split. Different products have different funnel shapes. A product with 3-month sales cycles may need 30/50/20; another with 6-month cycles may need 25/50/25.

3. W-Shaped Attribution

Principle: Extends position-based by adding explicit weight to lead creation stage[2].

Mathematical Structure: [ \text{First Touch} = 30% ] [ \text{Lead Creation} = 30% ] [ \text{Remaining Touches} = \frac{40%}{\text{Number of Remaining Touches}} ]

Use Case: Complex B2B funnels where lead creation (form submission, demo request) is a distinct conversion event. Recognizes that awareness → lead creation → deal closure are three distinct stages.

4. Data-Driven (Machine Learning) Attribution

Principle: Uses historical conversion patterns to dynamically assign credit based on actual customer behavior[2][4].

Mathematical Structure: [ \text{Credit}_i = P(\text{Conversion} | \text{Touchpoint}_i) - P(\text{Conversion} | \text{No Touchpoint}_i) ]

Where P(Conversion | Touchpoint) is the probability of conversion given that touchpoint occurred, and P(Conversion | No Touchpoint) is the baseline conversion probability without it.

Implementation: Requires sufficient historical data (typically 10,000+ conversions) and integration with a CDP or advanced analytics platform. Models analyze which touchpoint sequences predict conversion, then assign credit proportionally to each touchpoint's predictive power.

Advantage: Adapts to your specific funnel. If webinars drive 70% of conversion probability, the model assigns accordingly—not based on assumptions.

Limitation: Requires data science resources and 3–6 months of data collection before reliable output. Black-box nature can obscure how credit is assigned[4].

---

Platform-Specific Nuances for Paid Acquisition

Meta (Facebook/Instagram)

Meta's native attribution window: 28 days for clicks, 1 day for views. This creates systematic underweighting of Meta's true influence in longer sales cycles.

Optimization Strategy: Use position-based or time-decay models with custom windows (60–90 days for B2B). Meta's pixel data captures middle-funnel engagement; don't let their 28-day window artificially suppress budget allocation.

CAC Calculation: [ \text{Meta CAC} = \frac{\text{Meta Spend}}{\text{Conversions Attributed (60-day window)}} ]

TikTok

TikTok's strength: upper-funnel awareness and viral reach. Last-touch or linear attribution systematically undervalues TikTok because conversions typically occur on other platforms (Google, direct).

Optimization Strategy: Use position-based attribution with 50%+ weight on first touch. Measure TikTok's influence on downstream conversions (Google search volume, direct traffic) rather than direct conversions.

Metric: [ \text{TikTok Influence} = \frac{\text{Conversions with TikTok in Journey}}{\text{Total Conversions}} - \frac{\text{Conversions without TikTok}}{\text{Total Conversions}} ]

Google Search

Google Search's advantage: high-intent, late-funnel traffic. Last-touch attribution inflates Google's credit, but this is often justified—search captures users already in decision mode.

Optimization Strategy: Use time-decay or last-touch for search. Search's role is conversion, not awareness. Don't artificially suppress search budget by applying position-based models that split credit equally.

Reddit

Reddit's role: community engagement, niche awareness, and word-of-mouth amplification. Linear attribution underweights Reddit because conversions often occur after users leave Reddit and convert on other channels.

Optimization Strategy: Use position-based with 30%+ first-touch weight. Track Reddit users' downstream behavior (pixel data, UTM tracking) to measure influence on later conversions.

---

Scaling Frameworks: From Model Selection to Budget Allocation

Step 1: Audit Your Funnel

Map your actual customer journey:

• Awareness stage: Which channels introduce your brand? (TikTok, YouTube, Reddit, display)
• Consideration stage: Which channels nurture interest? (Email, retargeting, content)
• Decision stage: Which channels drive conversion? (Google Search, direct, sales calls)

Step 2: Select Attribution Model Based on Cycle Length

Step 3: Calculate Channel-Specific CAC

For each channel, calculate attributed CAC: [ \text{Channel CAC} = \frac{\text{Channel Spend}}{\text{Attributed Conversions (via selected model)}} ]

Step 4: Optimize Budget Allocation

Allocate budget proportionally to channels with lowest CAC relative to LTV:

[ \text{Budget Allocation}_i = \frac{\text{LTV} / \text{CAC}i}{\sum{j=1}^{n} (\text{LTV} / \text{CAC}_j)} \times \text{Total Budget} ]

Channels with LTV:CAC > 3:1 receive increased allocation. Channels with LTV:CAC < 1.5:1 receive reduced allocation or paused.

Step 5: Iterate Monthly

Attribution models should evolve as your funnel matures. Re-evaluate quarterly:

• Are attributed conversions matching actual revenue?
• Are high-CAC channels driving downstream LTV?
• Has your sales cycle length changed?

---

Common Pitfalls and How to Avoid Them

Pitfall 1: Assuming Linear Attribution is "Fair"

Linear attribution is not fair—it's mathematically convenient. Fairness requires weighting by actual influence, not equal distribution[4].

Solution: Audit your conversion data. Identify which touchpoints correlate with conversion. Weight accordingly.

Pitfall 2: Ignoring Cost-Per-Touchpoint

A channel that appears in 80% of journeys but costs 10x more per impression than competitors will have inflated attributed credit.

Solution: Calculate cost-per-attributed-conversion, not just attributed conversions. A channel with 100 attributed conversions at $10k spend (CAC $100) is worse than a channel with 50 attributed conversions at $2k spend (CAC $40).

Pitfall 3: Using Platform-Native Attribution Windows

Meta's 28-day window, Google's 30-day window, and TikTok's 1-day view window are not your actual sales cycle[4].

Solution: Use a unified attribution window (60–90 days for B2B, 14–30 days for e-commerce) across all platforms. Implement server-side tracking or CDP to enforce consistent windows.

Pitfall 4: Confusing Correlation with Causation

A touchpoint appearing in 90% of journeys doesn't mean it caused conversion—it may be a passive, habitual interaction[4].

Solution: Run incrementality tests. Pause a channel for 2 weeks and measure conversion impact. If conversions don't drop, that channel's attributed credit is inflated.

Pitfall 5: Not Accounting for Seasonality

Attribution models trained on Q4 data will misweight channels when applied to Q1. Seasonal shifts change funnel dynamics.

Solution: Retrain models quarterly. Segment data by season and adjust weights accordingly.

---

Actionable Next Steps

1. Audit current attribution: Document which model you're using and why. If it's linear, quantify the economic cost (estimated CAC inflation).

2. Map your funnel: Identify awareness, consideration, and decision stages. Determine average sales cycle length.

3. Select non-linear model: Choose position-based for most B2B, time-decay for short-cycle, data-driven for complex funnels.

4. Implement tracking: Ensure all touchpoints are captured with consistent timestamps and cost data. Use UTM parameters, pixel tracking, and server-side events.

5. Calculate channel CAC: For each channel, compute attributed conversions and CAC under your new model. Compare to LTV.

6. Reallocate budget: Shift spend toward channels with lowest CAC relative to LTV. Expect 15–30% ROAS improvement within 60 days.

7. Measure incrementality: Run holdout tests to validate that attributed channels actually drive conversions.

Linear attribution is a starting point, not a destination. Non-linear models require more infrastructure but deliver exponentially better unit economics. At scale, the difference between linear and data-driven attribution is the difference between guessing and optimizing.