Which traffic was worth buying, answered with what the wallets actually held

What was missing

Crypto marketers were buying traffic on clicks and sign-ups, with no way to see whether a source brought in wallets that held or moved real value.

What was built

Each marketing source was joined with on-chain transaction data for the wallets it brought in, then scored, segmented and sent back to the ad platforms. A schema-aware agent turns plain-language questions into BigQuery SQL, so the client's own team can ask the warehouse without writing SQL.

What changed

• 01Carried $4.5M of client media spend across its life
• 02Traffic judged on what it was worth, not on clicks
• 03Rebuilt across three stacks as the product matured