Challenge
An EU B2B marketing-analytics platform promised its clients one answer to the only question that matters in paid acquisition: which spend produced which deals? Under the hood, spend lived in three ad platforms, leads lived in a CRM, and the joins between them were manual. Every new client meant bespoke spreadsheet work, and attribution stopped at the click.
Approach
I was brought in to design the data platform from zero and deliver it in phases:
- A unified data model (UDM) covering campaigns, ad sets, creatives, UTM touchpoints, leads, and deals — one schema every client’s data lands in, whatever the source mix.
- A BigQuery medallion architecture: Bronze raw landings per connector, Silver conformed entities, Gold marts shaped for the product’s reports.
- Attribution models joining UTM-tagged touchpoints to CRM pipeline stages, so spend is credited against qualified leads and closed deals — not clicks.
- A connector strategy with per-source field mappings and a self-service path for adding new ad accounts without engineering time.
- A phased work breakdown — from a 90-hour first phase to a full 390-hour scope — so the client bought outcomes in increments, not a monolith.
- 01Discovery
- 02Requirements & data model
- 03Phase-1 build
- 04Scale-out
Result
- The platform’s five core client reports now run from Gold marts instead of hand-built sheets.
- Attribution reaches from ad spend to closed CRM deals, the exact promise the product makes.
- New data sources land as connector configs plus mappings — the medallion absorbs them without remodeling.