| name | attribution-modeling |
| description | Multi-touch attribution modeling for advertising campaigns. Use when analyzing conversion paths, comparing attribution models, implementing data-driven attribution, or understanding assisted conversions. |
Attribution Modeling
Analyze and implement multi-touch attribution models to understand how each marketing touchpoint contributes to conversions. Covers first-touch, last-touch, linear, time-decay, position-based, data-driven, and Markov chain attribution across Google Ads, Meta, and analytics platforms.
Capabilities
- Model Selection - Recommend the best attribution model based on sales cycle length, channel mix, and business goals
- Conversion Path Analysis - Map customer journeys from first interaction to conversion
- Assisted Conversion Valuation - Quantify the contribution of upper-funnel channels
- Cross-Channel Comparison - Compare how different models redistribute credit across channels
- Data-Driven Attribution - Implement algorithmic attribution using Markov chains or Shapley values
- GA4 Attribution Configuration - Set up and interpret GA4's data-driven attribution
Attribution Models Reference
1. First-Touch Attribution
Formula: 100% credit → first interaction
When to use:
- Brand awareness campaigns
- Top-of-funnel budget justification
- Short sales cycles (< 7 days)
Strengths: Simple, highlights discovery channels
Weaknesses: Ignores nurturing touchpoints, overvalues awareness
2. Last-Touch Attribution
Formula: 100% credit → last interaction before conversion
When to use:
- Direct response campaigns
- Short consideration windows
- Default model in most ad platforms
Strengths: Simple, clear ROI per channel
Weaknesses: Ignores awareness and consideration touchpoints
3. Linear Attribution
Formula: Credit = 1 / N (where N = total touchpoints)
Example: 4 touchpoints → each gets 25%
When to use:
- Equal-value touchpoint assumption
- Long sales cycles with consistent engagement
- B2B with multiple decision influencers
Strengths: Fair distribution, simple math
Weaknesses: Overvalues low-impact touches, no time weighting
4. Time-Decay Attribution
Formula: Credit_i = 2^((t_i - t_conversion) / half_life) / Σ 2^((t_j - t_conversion) / half_life)
Default half-life: 7 days (Google Ads default)
Example (14-day path, 7-day half-life):
| Touchpoint | Days Before Conversion | Raw Weight | Credit |
|---|
| Display Ad | -14 | 2^(-14/7) = 0.25 | 10.5% |
| Email | -7 | 2^(-7/7) = 0.50 | 21.1% |
| Search Ad | -3 | 2^(-3/7) = 0.74 | 31.2% |
| Brand Search | 0 | 2^(0/7) = 1.00 | 42.1% |
When to use:
- Promotions with deadlines
- Seasonal campaigns
- When recency matters most
5. Position-Based (U-Shaped) Attribution
Formula:
- First touch: 40%
- Last touch: 40%
- Middle touches: 20% / (N - 2)
Example (5 touchpoints):
| Position | Credit |
|---|
| First (Display) | 40% |
| Middle (Email) | 6.67% |
| Middle (Social) | 6.67% |
| Middle (Retarget) | 6.67% |
| Last (Brand Search) | 40% |
When to use:
- B2B lead generation
- When introduction and closing are most valuable
- Medium-length sales cycles (14-30 days)
6. Data-Driven Attribution (DDA)
How it works: Uses machine learning to analyze all conversion paths and assign credit based on statistical impact.
Requirements:
- Google Ads: 3,000 ad interactions + 300 conversions in 30 days
- GA4: Available for all properties (uses Shapley values)
Advantages:
- No assumptions about touchpoint value
- Adapts to actual customer behavior
- Accounts for path position, frequency, and sequence
Markov Chain Attribution
Concept
Models the customer journey as a state machine where each channel is a state with transition probabilities.
Step-by-Step Implementation
Step 1: Build the transition matrix
From conversion paths, calculate probability of moving from one state to another:
States: [Start, Paid Search, Display, Email, Social, Conversion, Null]
Transition Matrix P:
PS Display Email Social Conv Null
Start [0.35, 0.25, 0.20, 0.15, 0.00, 0.05]
PS [0.10, 0.05, 0.15, 0.05, 0.55, 0.10]
Display [0.30, 0.05, 0.20, 0.10, 0.15, 0.20]
Email [0.15, 0.10, 0.05, 0.05, 0.50, 0.15]
Social [0.20, 0.15, 0.15, 0.05, 0.25, 0.20]
Step 2: Calculate total conversion probability
P(conversion) = probability of reaching "Conversion" state from "Start" using matrix absorption.
Step 3: Calculate removal effect
For each channel, remove it from the model and recalculate conversion probability:
Removal Effect(channel) = 1 - P(conversion without channel) / P(conversion with all channels)
Step 4: Normalize to get attribution weights
Attribution(channel) = Removal Effect(channel) / Σ Removal Effect(all channels)
Example Calculation
| Channel | P(conv) with all | P(conv) without | Removal Effect | Attribution |
|---|
| Paid Search | 0.42 | 0.18 | 57.1% | 38.2% |
| Display | 0.42 | 0.35 | 16.7% | 11.2% |
| Email | 0.42 | 0.22 | 47.6% | 31.9% |
| Social | 0.42 | 0.30 | 28.6% | 19.1% |
Workflows
Workflow 1: Model Comparison Analysis
-
Pull conversion paths from GA4:
- Navigate to Advertising → Attribution → Conversion paths
- Set date range to 30+ days
- Filter by conversion event
-
Export path data with touchpoint sequence, timestamps, and conversion values
-
Apply each model to the same dataset:
- Calculate revenue attribution per channel per model
- Create comparison table showing credit shifts
-
Identify discrepancies:
- If Channel A gets 40% on last-touch but 15% on first-touch → it's a closer, not a discoverer
- If Channel B gets 5% on last-touch but 30% on first-touch → it's an awareness driver
-
Recommend model based on business stage:
- Growth stage → Position-based or DDA
- Efficiency stage → Time-decay or DDA
- Brand building → First-touch or DDA
Workflow 2: GA4 Attribution Setup
-
Access attribution settings:
- Admin → Attribution Settings
- Select reporting attribution model (default: Data-driven)
- Set lookback window: 30 days for acquisition, 90 days for other conversions
-
Configure conversion events:
- Admin → Events → Mark as conversion
- Set counting method: Once per session (leads) or Every event (purchases)
-
Review attribution reports:
- Advertising → Attribution → Model comparison
- Compare DDA vs Last-click for each channel
- Look for channels where DDA credits significantly more than last-click
-
Export and act:
- Increase budget on channels undervalued by last-click
- Decrease budget on channels overvalued by last-click
- Monitor for 2-4 weeks before further adjustments
Workflow 3: Assisted Conversions Analysis
-
Pull assisted conversion data from GA4:
- Advertising → Attribution → Conversion paths
- Note "Assisted conversions" vs "Last-click conversions" per channel
-
Calculate Assisted/Last ratio:
Assist Ratio = Assisted Conversions / Last-Click Conversions
-
Interpret ratios:
| Ratio | Interpretation |
|---|
| < 0.5 | Primarily a closer (last-touch channel) |
| 0.5 - 1.0 | Balanced role (both assists and closes) |
| 1.0 - 3.0 | Primarily an assister (mid-funnel) |
| > 3.0 | Almost exclusively awareness/discovery |
-
Budget implications:
- High-assist channels deserve more budget than last-click ROAS suggests
- Cutting high-assist channels will reduce conversions on other channels
Reference Data
Typical Attribution Shifts by Channel
| Channel | Last-Touch Credit | DDA Credit | Typical Shift |
|---|
| Brand Search | 35-45% | 20-30% | ↓ Overvalued |
| Non-Brand Search | 20-30% | 20-25% | → Stable |
| Display | 3-8% | 8-15% | ↑ Undervalued |
| Social (Organic) | 5-10% | 8-12% | ↑ Slightly undervalued |
| Social (Paid) | 8-15% | 10-18% | ↑ Slightly undervalued |
| Email | 10-20% | 8-15% | ↓ Slightly overvalued |
| Direct | 15-25% | 10-15% | ↓ Overvalued |
Lookback Window Recommendations
| Business Type | Recommended Window | Rationale |
|---|
| E-commerce (low AOV) | 7-14 days | Impulse purchases |
| E-commerce (high AOV) | 30 days | Research phase |
| B2B SaaS | 60-90 days | Long sales cycle |
| B2B Enterprise | 90+ days | Multiple stakeholders |
| Local services | 7-14 days | Urgency-driven |
| Travel / Hospitality | 30-45 days | Planning phase |
Platform-Specific Attribution Defaults
| Platform | Default Model | Lookback (Click) | Lookback (View) |
|---|
| Google Ads | Data-driven | 30 days | 1 day |
| Meta Ads | 7-day click, 1-day view | 7 days | 1 day |
| LinkedIn Ads | Last-touch | 90 days | 7 days |
| TikTok Ads | Last-touch | 7 days | 1 day |
| Reddit Ads | Last-touch | 28 days | 1 day |
Examples
Example 1: B2B SaaS Attribution Analysis
Scenario: A SaaS company spends $50K/month across Google, LinkedIn, and content marketing. Last-click shows Google at 4x ROAS and LinkedIn at 0.8x ROAS.
Analysis with position-based model:
- LinkedIn drives 45% of first touches (awareness)
- Google captures 60% of last touches (closing)
- Content marketing handles 70% of middle touches (nurturing)
Recommendation: Don't cut LinkedIn. It feeds the top of funnel. Cutting it would reduce Google's closing conversions within 30-60 days.
Example 2: E-commerce Markov Chain Results
Scenario: Online retailer with 10K conversion paths analyzed.
Removal effects:
- Removing Paid Social → 34% fewer conversions
- Removing Paid Search → 28% fewer conversions
- Removing Email → 22% fewer conversions
- Removing Display → 16% fewer conversions
Budget reallocation: Shift 15% of display budget to paid social, which has higher marginal impact per dollar despite lower last-click ROAS.
Example 3: Cross-Platform Attribution Conflict
Scenario: Meta reports 500 conversions, Google reports 400 conversions, but GA4 only shows 350 total conversions.
Root cause: Platform self-attribution double counts conversions where users saw both a Meta ad and clicked a Google ad.
Fix: Use GA4 DDA as source of truth. Apply platform-specific credit:
- Meta DDA credit: 210 conversions (60%)
- Google DDA credit: 140 conversions (40%)
- Total matches GA4: 350
About this skill
Part of the Synter free skills collection — open-source agent skills for advertising, PPC, and marketing automation.
Want this skill (and 40+ more) running on autopilot against your live ad accounts? Try Synter — AI Agent Media Buyers that connect to Google, Meta, LinkedIn, TikTok, Reddit, Amazon, and 7+ more platforms.
