Industry Architect · Adtech & Data Infrastructure

Thirty adtech case studies.
One practitioner's read on each.

I've spent two decades designing data systems at scale — from backend engineering through data platform architecture to leading AI product. These case studies cover the real architectural decisions behind adtech's hardest problems: why a team chose ClickHouse over Snowflake for sub-second OLAP, how Druid and Pinot handle streaming ingestion differently, where Databricks wins and where it doesn't.

This isn't survey content. It's what I actually think, informed by having made — and watched others make — these choices under real constraints.

Browse all 30 cases About this project

Featured cases

View all 30 →
Braze ClickHouse

Braze — Ads Analytics Pipeline Rebuild

Braze replaced a dual MongoDB + Snowflake pipeline with ClickHouse Cloud, achieving 8x performance improvement while handling 1B+ events per hour.
Reddit Apache Druid

Reddit — Ad Budget Pacing in 30ms

Reddit solved sub-30ms ad budget pacing with Apache Druid, achieving 100% accuracy even during Kafka and Flink outages — replacing failed batch and streaming-to-batch approaches.
Moloco Apache Pinot

Moloco — DSP: 10 Minutes → Milliseconds

Moloco cut query latency from 10 minutes to milliseconds by migrating its $1B+ ad platform to Apache Pinot + StarTree, now serving 6M queries/second.

About this site

This is a working document, not a polished presentation. Thirty adtech companies, each with a real architectural story — streaming OLAP trade-offs, Snowflake vs. ClickHouse cost curves, clean room implementations, MMM revival patterns. I read the case material, applied twenty years of systems context, and wrote down what I actually think. The "my angle" section on each case is the part worth reading. The rest is scaffolding.