Tasman Analytics released tasman-dbt-mta as an open-source dbt package after stress-testing it in production for several years. The package implements multi-touch attribution as a configurable, deterministic attribution engine.
Supported attribution models: last-touch, first-touch, and U-shaped (40% weight to first and last touches, 20% split across intermediary touches). Attribution windows are configurable per model (typically 7–90 days depending on industry).
Architecture: Two primary output models — attributed_touches (all filtered touches across all attribution models attributed to a conversion, with per-touch conversion share values) and attributed_conversions (conversion-level rollup). Touch and conversion rules are parameterized rather than hardcoded, enabling adaptation to different business logic without modifying package internals.
Supported targets: Snowflake and BigQuery.
Repository: https://github.com/TasmanAnalytics/tasman-dbt-mta
Note: Tiers 1 (github.com direct) and 2 (Wayback) were unavailable from this runtime environment. Content sourced from Tasman Analytics' own news page (Tier 3 alternative).