The Challenge
GlowCo (name changed) is a DTC skincare brand doing $40M in annual revenue. They were spending $2M per month across Meta, Google, TikTok, and influencer partnerships—but had no idea which channels actually drove purchases.
Each platform claimed credit for every conversion. Meta said they drove 80% of sales. Google said 70%. The numbers didn't add up, and the marketing team was flying blind.
Key Problems
- Platform over-reporting: Each ad platform claimed more conversions than actually occurred
- No cross-device tracking: Mobile browsers and app installs weren't connected
- iOS 14.5 blindness: 60% of traffic was from iOS with limited tracking
- Last-click tunnel vision: GA4 only showed last-click, missing the full journey
The Solution
We built a first-party data infrastructure using Jitsu for event collection and ClickHouse for the attribution warehouse.
Architecture Overview
- Jitsu: First-party event collection with server-side tracking
- ClickHouse: Attribution data warehouse with custom multi-touch models
- Conversion APIs: Server-side events to Meta, Google, and TikTok
- Metabase: Marketing performance dashboards
Attribution Model
We implemented a position-based attribution model with time decay: 40% credit to first touch (awareness), 40% credit to last touch (conversion), 20% distributed across middle touches with time decay. This was compared against linear and data-driven models to validate findings.
Implementation Process
Week 1: Data Audit and Identity Resolution
We mapped all touchpoints and built an identity graph connecting email, device IDs, and first-party cookies. This recovered 40% of previously "unknown" users.
Week 2: Jitsu Setup and Conversion APIs
We deployed Jitsu for server-side event collection, bypassing browser limitations. We set up Conversion API integrations with Meta, Google, and TikTok for server-side attribution.
Week 3: ClickHouse Attribution Models
We built attribution models in ClickHouse, joining ad spend data with conversion events. We created customer journey analysis showing typical paths to purchase.
Week 4: Dashboards and Training
We built marketing dashboards showing attributed revenue by channel, campaign, and creative. We trained the team on interpreting multi-touch data.
Results
Within 90 days, GlowCo had actionable insights that transformed their marketing strategy:
- 23% ROAS improvement: By understanding true channel performance
- $500K reallocation: Shifted budget from over-credited to under-credited channels
- TikTok discovery: Found TikTok was driving 3x more first-touch value than reported
- Influencer clarity: Finally measured influencer impact on downstream conversions
Key Insight
The biggest surprise was branded search. Google Ads was claiming massive conversion volume, but multi-touch analysis showed most of those customers had already decided to buy—branded search was just the final click. Reallocating that budget to upper-funnel channels drove net new customers.
"We were basically throwing money at channels based on their own reporting. Now we have a source of truth that actually tells us where our customers come from. The Jitsu + ClickHouse stack paid for itself in the first month."