Analytics Attribution

Marketing Attribution Modeling

When to Activate

• Allocating or reallocating marketing budget across channels
• Evaluating which campaigns or channels drive conversions
• Building or improving marketing measurement infrastructure
• Assessing impact of iOS/privacy changes on tracking
• Setting up UTM tracking and attribution tooling
• Debating "what's working" with stakeholders who disagree
• Running incrementality tests to validate attribution data

First Questions

1. What is your current attribution model and tooling? (GA4, platform pixels, MTA vendor, MMM?)
2. What does your conversion funnel look like? (Awareness -> consideration -> purchase -> retention)
3. How long is your typical customer journey? (Same-day impulse vs. 90-day B2B sales cycle)
4. What channels are you running? (Paid search, paid social, organic, email, direct, referral, affiliate)
5. What is your primary conversion event? (Purchase, sign-up, demo request, app install)
6. How much of your traffic is mobile vs. desktop? (Privacy impact assessment)
7. Do you have a CRM or CDP connecting touchpoints to customers?

Core Attribution Models

Last-Click Attribution

• How it works: 100% credit to the final touchpoint before conversion.
• Best for: Direct-response campaigns, short purchase cycles, bottom-of-funnel optimization.
• Limitation: Ignores all awareness and consideration touchpoints. Massively over-credits branded search and retargeting.
• When to use: As a baseline only. Never as your sole model.

First-Click Attribution

• How it works: 100% credit to the first touchpoint in the journey.
• Best for: Understanding top-of-funnel channel effectiveness, awareness campaigns.
• Limitation: Ignores everything that happens after initial discovery.
• When to use: When evaluating demand generation and awareness investments.

Linear Attribution

• How it works: Equal credit distributed across all touchpoints.
• Best for: When you genuinely believe every touchpoint matters equally.
• Limitation: Treats a random display impression the same as a high-intent search click.
• When to use: Early-stage attribution when you lack data for more sophisticated models.

Time-Decay Attribution

• How it works: More credit to touchpoints closer to conversion, decaying backward.
• Best for: Longer sales cycles where recent interactions are more influential.
• Limitation: Under-credits awareness touchpoints that may have been essential.
• When to use: B2B with 30-90 day sales cycles. E-commerce with multi-session journeys.

Position-Based (U-Shaped) Attribution

• How it works: 40% to first touch, 40% to last touch, 20% distributed across middle.
• Best for: Balanced view that values both discovery and closing channels.
• Limitation: Arbitrary weighting. Middle touches may matter more than 20%.
• When to use: Good default for most businesses. Balances awareness and conversion.

Data-Driven Attribution (DDA)

• How it works: Uses machine learning to assign credit based on actual conversion patterns.
• Best for: High-volume businesses with sufficient conversion data.
• Limitation: Requires significant data volume (GA4 needs 600+ conversions in 28 days). Black box — hard to explain.
• When to use: When you have the data volume. GA4 DDA is now default and accessible.

Model Selection Framework

| Factor | Recommended Model | |--------|-------------------| | Short sales cycle (20% of channel spend).

• When entering a new channel and need to validate early results.
• Quarterly on your top 2-3 spend channels.

Quality Gate

Before finalizing attribution analysis or recommendations:

• [ ] Have you compared at least two attribution models side by side?
• [ ] Have you accounted for cross-device and cross-platform journeys?
• [ ] Have you noted known tracking gaps (consent rates, iOS impact, ad blockers)?
• [ ] Are UTM parameters consistent and properly deployed?
• [ ] Have you checked for duplicate conversion counting across platforms?
• [ ] Have you complemented or planned to complement with incrementality testing?
• [ ] Are budget recommendations presented as directional, not falsely precise?
• [ ] Have you documented model assumptions and limitations for stakeholders?

Source & license

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

• Author: brainbytes-dev
• Source: brainbytes-dev/everything-claude-marketing
• License: MIT

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