How Hulu’s Segmentation System Powers Big Data Marketing at Scale

Speaker Background

Zhao Kunliang graduated from Tsinghua University Computer Science, has six years of internet development experience, and leads Hulu Software’s Beijing big data team.

Event Overview

The WOT 2016 Big Data Technology Summit was held on November 25‑26, 2016 in Beijing. The conference series, started in 2012, focuses on technology and serves IT professionals.

Presentation

At the “Industry Application Practice” session, Zhao delivered the talk “Hulu Software’s Big Data Marketing Practice”.

Segmentation System

Hulu’s big data development uses mainstream open‑source technologies. The underlying infrastructure relies on a Hadoop cluster, batch processing is handled by Spark, and streaming by Spark Streaming. Hulu also built its own query engine, Nesto, for the Segmentation system.

The Segmentation system supports Hulu’s marketing and advertising operations, providing precise data‑driven marketing and analytics, and is a core product of Hulu’s big‑data initiatives.

Challenges and Innovations

As data volume grew and requirements diversified, the Beijing big data team continuously optimized the Segmentation system. They created a JSON‑based DSL query language and a template tool that lets users build queries via drag‑and‑drop. To meet fast interactive queries on massive nested data, they enhanced Nesto for efficient execution. The system also integrates third‑party data and ad‑company data, enabling comprehensive marketing plan creation and performance evaluation.

About Hulu

Founded in 2008 in the United States, Hulu established its Beijing team from the start. The Beijing team collaborates closely with the US team, sharing development responsibilities across all product modules, embodying Hulu’s “One Team” culture.