From Gut Feel to Pipeline Clarity

Overview

The agency was growing, winning impressive clients, and spending on two things they couldn't measure: marketing and compute. €8k/month across paid channels and events, and €5k/month on GPU render farm infrastructure — with no clear accountability on either. Every lead source lived in a different tool. Every render job ran without a cost tag. The business was profitable on paper but completely blind on margin. Nobody could answer the two questions that actually matter: what does it cost to win a client, and how much is that client worth over time?

The Challenge

With studios across Berlin, Dubai, Cairo, and Paris, the BD team was making marketing budget decisions based on instinct — "LinkedIn feels like it's working" — with no closed-loop data to back it up. There was no CAC tracking by channel, no LTV model, and no way to connect a signed contract back to the campaign that started the conversation. The agency didn't know if their average client paid back their acquisition cost in month three or month twelve. They didn't know which channels produced the highest-value long-term clients versus which ones brought in one-project accounts that never returned. Meanwhile the infrastructure problem was equally expensive: GPU server costs arrived as one flat monthly line item with zero attribution. Two black boxes, €13k/month, zero accountability.

The Solution

I built a unified data layer connecting marketing, CRM, and infrastructure into one operational view:

• CRM Data Pipeline (Pipedrive + BigQuery):: I pulled the full deal history — stages, sources, owners, values, close dates, and contract lengths — into BigQuery across all four offices. One clean, queryable record of every opportunity the agency had ever touched.
• CAC by Channel (dbt):: Built attribution models connecting marketing spend across Google Ads, LinkedIn, events, and outbound to closed deals. True Customer Acquisition Cost per channel — not just cost-per-lead, but cost-per-signed-contract.
• LTV Modelling:: Built a client Lifetime Value model based on average project value, repeat engagement rate, and contract length by client segment. Stacked against CAC, this revealed which channels were producing the agency's best long-term clients versus one-off transactional wins.
• Pipeline Health Dashboard (Looker Studio):: Live view of open deals by stage, weighted pipeline value, lead-to-close conversion rate, average sales cycle by channel, and win rate by deal source — updated daily.
• GPU Cost Attribution (Python + BigQuery):: Render job logs tagged to client project IDs, surfacing which projects were burning the most compute and which were silently re-running failed jobs.
• Unified Margin View:: Marketing spend, GPU infrastructure costs, and invoiced revenue in one dashboard. Leadership could finally see true project margin — and whether the CAC paid to win a client was actually justified by what that client delivered over time.