E-Commerce Data Lakehouse

The Challenge

Product, marketing, and finance teams each maintained their own exports from the e-commerce platform, ad networks, and payment processor. The numbers never quite matched between teams, and every new report request meant another one-off SQL script bolted onto an already fragile setup.

The Approach

The fix was to stop exporting and reconciling, and instead give every team one shared, governed source to query.

• Used Fivetran connectors to automatically replicate the e-commerce platform, ad spend, and payments data into Snowflake.
• Modeled a conformed dbt layer with shared definitions for revenue, orders, and customer acquisition cost.
• Exposed curated marts through Looker so teams could explore data without writing SQL.
• Set up incremental models and freshness checks so dashboards always reflected the latest day's activity.

The Results

Three overlapping spreadsheets became one lakehouse that marketing, product, and finance all trusted equally.

Key Takeaways

Most of the "data doesn't match" friction disappeared the moment revenue and CAC had one shared dbt definition instead of three slightly different spreadsheet formulas. Managed ingestion freed up time to focus on the modeling layer, which is where the real value was.