Marc.Turtal
  1. Home
  2. Work
  3. Revenue warehouse
Case study 1 of 4 · 2026 · A high-ticket sales education company, Australia

From zero to a revenue source of truth in 8 weeks.

Eight tools, no attribution, and nobody could say how much cash a month actually collected. One warehouse fixed that, built solo.

  • 21 salesFully tracked in the first complete month.
  • A$207kRevenue visible end to end, by channel, matching the client's public figures 1:1.
  • A$82.5kCash collected, reconciled against payments.

Problem

Revenue data scattered across 8 tools, no attribution. Nobody could say which channel produced which sale, or how much cash a month actually collected.

Build

A Supabase Postgres warehouse on a three-layer architecture, fed by 12+ integrations: CRM, accounting, payments, paid ads, Instagram DM sales, Messenger, Discord, YouTube, call recordings, video analytics, webinars, and an intake form pipeline.

The warehouse now also carries paid spend and cash: a Meta Ads pipeline verified to the cent, and a payments feed tracking A$1.07M+ in running cash.

A read-only MCP server lets executives query the warehouse in plain English in Claude. Everything runs unattended on Railway crons.

The system, end to end
12+ sources CRM, GoHighLevel Accounting, Xero Payments Paid ads, Meta Instagram DMs Messenger, Discord YouTube, Webinars Call recordings Video analytics Intake forms Postgres warehouse Supabase, versioned migrations STAGING raw, append only CORE typed, entity resolved LOGIC governed views, read only MCP server SELECT only, on Railway Executive in Claude asks in plain English no SQL required CRM, GoHighLevel Accounting, Xero Payments Paid ads, Meta Instagram DMs Messenger, Discord YouTube, Webinars Call recordings Video analytics Intake forms Postgres warehouse Supabase, versioned migrations STAGING raw, append only CORE typed, entity resolved LOGIC governed views, read only MCP server SELECT only, on Railway Executive in Claude asks in plain English 12+ sources one source of truth no SQL required

Stack in this build

Supabase Postgres Claude + MCP Railway Discord YouTube Instagram DMs Messenger

Delivery

180+ PRs, delivered solo, AI-native. 27,500+ Discord messages captured, starting from a 14,297-message backfill across 26 channels. 743 YouTube videos pipelined.

The write-up · 8 min readShipping a revenue warehouse solo in 8 weeks Read the article

Want this for your sales team?

Most engagements start with attribution or a warehouse and grow from there.

See the revenue data offer