Marketing attribution is the practice of assigning conversion credit to the marketing touchpoints that preceded it. Every approach involves tradeoffs — between simplicity and accuracy, between implementation effort and insight depth, between what’s measurable and what actually drives business outcomes.
This hub organizes the attribution analysis journey from first principles through production implementation.
The Sequence
1. Understand the measurement problem
Ad Platform Attribution Bias explains why platform-reported attribution numbers can’t be trusted at face value. Every ad platform over-counts conversions due to walled-garden incentives, overlapping attribution windows, and the inability to see cross-channel journeys. This is the problem that warehouse-based attribution solves.
2. Assemble the data sources
Warehouse Attribution Data Sources covers the three categories of data required for attribution: website interaction data (GA4 BigQuery export), ad platform spend data (Google Ads DTS, Meta, LinkedIn, TikTok via ETL), and conversion data (e-commerce, CRM, product events). Includes the GA4 gclid misattribution bug, join strategies between ad data and web analytics (click IDs vs. UTMs), and UTM standardization rules.
3. Design the touchpoint table
Attribution Touchpoint Table Design covers the foundational data structure that every attribution model consumes. The touchpoint table joins marketing interactions to conversions, and its design determines attribution accuracy. Includes required fields, the identity resolution challenge, what counts as a touchpoint, the intermediate dbt model that builds it from GA4 events, and common mistakes.
4. Set your lookback windows
Attribution Lookback Windows covers how far back to consider touchpoints before a conversion. Industry benchmarks, consequences of too-short and too-long windows, platform default comparisons, SQL implementation, and how to use your own conversion data to find the right window.
5. Implement attribution models in SQL
SQL Attribution Patterns covers the SQL implementation for every major attribution model: first-touch, last-touch, linear, position-based, time-decay, and algorithmic models (Markov chains, Shapley values). Each model operates on the same touchpoint table and differs only in how it distributes credit. The note includes the touchpoint table schema, full SQL for each model, and a decision framework for choosing models.
6. Go deeper on weighted models
Position-Based Attribution Models covers the U-shaped (40-20-40) and W-shaped (30-30-30-10) models in detail — the formulas, edge case handling, industry weight variations, and BigQuery SQL. Time-Decay Attribution Model explains exponential decay with a configurable half-life, how to choose the right half-life for your sales cycle, and the normalization math that keeps attributed revenue balanced.
dbt Weighted Attribution Models shows how to implement both in dbt with configurable weights via var(), including dbt_project.yml configuration and revenue integrity tests.
7. Group channels for data-driven models
Attribution Channel Grouping Strategy covers how to group marketing channels before running data-driven models. Start with 5-10 high-level groups, collapse low-volume channels (under 2% of touchpoints) into “Other,” and adjust groupings quarterly. Includes the impact on Markov transition matrices vs. Shapley coalition counts, the 2% rule, and iteration cadence.
8. Implement data-driven attribution
Markov Chain Attribution explains how Markov chains model customer journeys as state transitions to calculate attribution through the removal effect — how much conversion probability drops when a channel is removed. Covers the Markov property, transition matrices, first-order vs. higher-order models, and data requirements.
Markov Attribution SQL Implementation provides the BigQuery SQL for path extraction and transition counting — the data preparation layer that feeds into Python for removal effect computation. Includes the dbt project structure for integrating Markov results alongside heuristic models.
Shapley Value Attribution covers the game theory alternative: calculating each channel’s average marginal contribution across all possible channel coalitions. Shapley values satisfy four fairness axioms (efficiency, symmetry, dummy player, additivity) but scale at 2^n for n channels. Includes the Shapley vs. Markov decision framework.
9. Understand Google DDA’s limitations
Google DDA Silent Fallback reveals a critical trap: GA4’s Data-Driven Attribution silently falls back to last-click when data thresholds aren’t met (400+ conversions per type per 30 days, ~10,000 multi-interaction paths). The interface still says “Data-Driven Attribution” while delivering last-click. Warehouse-native attribution avoids this entirely.
10. Evaluate dbt packages and libraries
dbt Attribution Packages Landscape surveys the open-source dbt packages (Snowplow attribution, Snowplow fractribution, Tasman MTA, Rittman Analytics) and Python libraries (ChannelAttribution, marketing-attribution-models) available for production attribution. Covers what each includes, warehouse support, and the build-vs-package decision framework.
11. Build the comparison layer in dbt
dbt Attribution Comparison Pattern shows how to structure a dbt project for multi-model attribution. Each model runs independently as a mart, and a comparison model unions them together with a model_type discriminator. Includes the project structure, the comparison model SQL, revenue integrity tests, schema contracts, and pre-aggregation patterns for dashboard performance.
12. Interpret model disagreement
Attribution Model Disagreement as Signal covers the core insight: the disagreement between attribution models is itself informative. When models agree, confidence is high. When they disagree, the spread reveals genuine uncertainty in your data. Includes the disagreement score calculation, strategy alignment across teams, stakeholder communication frameworks, and warning signs that attribution is failing.
13. Triangulate with multiple methodologies
Triangulated Marketing Measurement covers why resilient measurement combines three approaches: multi-touch attribution for daily digital optimization, media mix modeling (MMM) for quarterly strategic budget allocation, and incrementality testing for causal validation. Includes how each approach calibrates the others and a maturity-based roadmap for adopting each methodology.
14. Design attribution dashboards
Attribution Dashboard Design covers dashboard architecture for multi-model attribution: essential metrics (channel contribution, model comparison, assisted conversions, CPA/ROAS), audience-tiered hierarchy (executive/manager/analyst views), Looker Studio implementation, and working around BI tool limitations.
15. Validate with incrementality testing
Incrementality Testing for Attribution closes the loop. Attribution measures correlation; incrementality tests measure causation. Covers holdout tests, geo tests, platform lift studies, and how to use incrementality results to calibrate your attribution model interpretation.