Customer Data Platform
21M+ unique consumer profiles. 1B+ behavioral data points. 6 DSP integrations. Activated by a major DSP partner for pharmaceutical advertiser campaigns. The CDP was the connective tissue between the consumer ordering app and advertiser activation - the layer that turned ordering behavior into programmable audiences.
Six layers, one platform
The CDP wasn't a vendor purchase - it was built in-house. Every layer designed to do one thing well and hand off cleanly to the next. The whole thing sits between the consumer app and the DSPs, turning raw behavior into segments advertisers can target.
Identity Stitching Layer
Cross-device, cross-event consumer matching. Deterministic where possible (hashed PII), probabilistic for the rest. Resolves anonymous browsers to known users as data accrues.
Pixel Event Capture
Custom JavaScript pixel firing across the consumer app. Page views, product views, dwell, cart, checkout - everything captured in real time, schema-versioned.
DSP Activation Pipeline
Audience definitions translate into DSP-native segment formats. Real-time and batch sync. Lookalike modeling on top of seed audiences. Suppression and frequency capping handled server-side.
Real-Time + Batch Sync
Hot audiences refresh in minutes; lookalikes and broad segments refresh nightly. Six DSP destinations, each with its own quirks and rate limits, normalized behind one interface.
Compliance Layer
Built for regulated-industry advertising. Consent management, audience eligibility rules, restricted-category gating. The reason the platform could be activated for pharmaceutical campaigns.
How profiles get built
Anyone can capture an event. The hard part is turning a billion isolated events into 21 million coherent profiles a DSP can target. That's the identity stitching layer.
- First-party signals from app.jointcommerce.com - signup, browse, add-to-cart, checkout, repeat purchase
- Deterministic matching where possible - hashed PII (email, phone) where consent allows
- Probabilistic stitching for the rest - device, session, IP, and behavior fingerprints
- Result: 21M+ unique profiles and 1B+ behavioral data points in a single addressable graph
Six DSPs. One pipeline.
Six direct DSP integrations at JointCommerce. Across my career I've worked across DV360, TTD, Xandr, Yahoo, Zeta, Basis, Quantcast, and Vistar - so the activation pipeline was designed for portability from day one.
- Audience export pipeline to 6 DSPs - normalized API layer translates one segment definition into six native formats
- Lookalike modeling on top of seed audiences - first-party seeds extended into broader prospect pools
- Suppression and frequency capping handled at the CDP level so DSPs receive clean, deduplicated audiences
- ICP-matched segments - segments built around the actual ideal customer profile, not vendor presets
- Audience freshness SLAs - hot segments refresh in minutes, broad lookalikes refresh nightly
The pharmaceutical activation
The clearest external validation of the platform was a major DSP partner activating the CDP for a regulated-industry use case where most data platforms can't participate.
A major DSP partner activated the CDP for pharmaceutical campaigns
A major DSP partner activated the CDP for pharmaceutical advertiser campaigns. Pharma is one of the most heavily regulated advertising verticals in the US, with specific compliance requirements - FDA direct-to-consumer rules, restricted indication targeting, and HIPAA boundaries where applicable.
Most CDPs and audience platforms can't be activated for pharma because they weren't built with a compliance layer in mind. Our CDP was. The regulated-industry framework baked into the architecture - consent management, audience eligibility rules, restricted-category gating - was what made the activation possible.
Access to high-quality, first-party purchase-intent audiences with verified attribution and a defensible compliance posture. Real signal, not modeled.
DSP partner revenue from a vertical most competitors can't touch. First-party data monetized through a media partnership rather than direct sale.
The methodology
The numbers are specific to JointCommerce. The methodology is portable. Here's the parts I'd carry into any next role.
Two-sided platform thinking
The consumer side generates data; the advertiser side activates it. The two only work together. Build them as one platform, not two products.
Identity stitching as a first-class layer
Identity isn't a feature - it's the foundation. First-party signals, deterministic where possible, probabilistic for the rest.
DSP partnerships beat direct sale for activation
Selling data direct is hard. Letting a DSP partner activate your data and split revenue is faster, scales further, and unlocks regulated verticals the direct-sale motion can't reach.