Alex Clark

At Amazon, I designed and implemented a multi-touch attribution (MTA) pipeline to assign credit to individual pages for purchases, complementing the existing content-level attribution model. The pipeline processed billions of rows across web hits, orders, products, and page-hierarchy tables, handling complex relationships, including ASIN variants, browse-tree hierarchies, and multi-item orders. I developed logic to ensure relevance, assigning credit only to products linked to the pages visited within a 24-hour window. The resulting analytics datasets included CTR, retention, and page-level MTA metrics, enabling product teams to comprehensively measure page performance. The pipeline was partitioned by marketplace and date, optimized for large-scale data in AWS EMR and Athena, and provided a repeatable, reliable framework for ongoing page-level analysis.

Led the design and implementation of an internal API-based billing system to track platform usage and allocate infrastructure costs. Collaborated with cross-functional teams to identify cost drivers across multiple services and developed a dynamic pricing model that accounted for compute, storage, and API usage. Built end-to-end data pipelines and analytics dashboards to support transparency, reporting, and operational decision-making, delivering the system on time while ensuring accuracy and scalability.

At Amazon, I designed and implemented a scalable ETL pipeline for the Content Platform, which handles images, videos, widgets, and other retail site content across four regions (NA, EU, CN, FE). I consolidated nine versioned (Type 2) source tables into analytics-optimized, regionally partitioned Parquet datasets with Snappy compression. Each day, the pipeline automates ingestion, pivoting of key-value placement properties, version selection, and joins, producing one denormalized table with the most frequently queried fields. This enabled teams to quickly answer questions about impressions, click-through rates, placement attribution, and content performance, without querying multiple regions or spinning up expensive Redshift resources. The solution processed hundreds of terabytes of JSON data via AWS EMR, Hive, and Athena, and was eventually integrated into the Data Warehouse. This project improved query performance, reduced operational complexity, and made large-scale content analytics accessible to product, analytics, and business teams.