Proof
What we replaced. Client names shared with permission, anonymized otherwise.
US performance agency – context warehouse
type: Data warehouse + multi-client dashboard + AI-ready foundation
what was replaced: Reporting scattered across platform UIs and spreadsheets, no cross-channel view, no historical data beyond what the platforms retain
how: BigQuery warehouse with daily pipelines through the official APIs (Google Ads, Meta, Northbeam, AppLovin, Pinterest, Shopify, Triple Whale), historical backfill, medallion architecture. Custom dashboard on top: business health, channel performance, campaign and creative breakdowns, across every client. Operations registry structured so an LLM can query the right tables instead of drowning in raw rows.
result: Every client, every channel, one place. History preserved. The foundation their AI layer runs on.
status: Live
vendezvitevendezbien.fr
type: Lead generation platform + marketing automation
what was replaced: Manual lead generation — referrals, cold prospecting, no digital pipeline
how: Built a real estate price analysis tool on French government transaction data (DVF). Visitors explore property prices for Southwest France, then hit a lead gate after 3 interactions. Email captured to Supabase, triggers a 4-email Brevo nurture sequence, pushes toward a property estimation request. Meta ad campaigns to drive traffic to the lead magnet. PPC AI Agent via Markifact to manage, optimise and report on campaigns.
result: Fully automated lead pipeline. Warm leads delivered by email with no manual follow-up setup.
status: Live
contact: Naum Durieux
15 years of paid media from the inside. We know which click-work costs the most, and what it takes to make the automation actually stick.