Context

Pitch's attribution model had overlapping channel definitions. Sessions were being assigned to the wrong channels, or falling into a catch-all "other" bucket. Marketing spend decisions were being made on numbers that didn't reflect reality.

Scope

Redesign the channel definitions and attribution logic. Make it accurate, granular, and automatic.

Approach

We redefined 8 marketing channels and introduced 19 sub-channels for finer-grained insight into what was actually driving signups. Two new data sources were added to Snowflake via Airbyte. The attribution model was built in dbt — one new model, seven updated — and runs retroactively, so historical data was reclassified too. To prevent future drift, we built a UTM generator in Google Sheets so marketers could produce correctly-formatted links automatically.

Signup attribution accuracy by marketing channel

Outcome

The model now accurately attributes 70% of signups to a specific channel. Attribution accuracy improved by 24%. The "other" bucket shrank from 10% to 1%. Pitch could finally see which acquisition channels were bringing in users who convert and pay — not just sign up.