Building an Effective Data Platform: Insights and Practices from Tencent's Senior Product Manager

In this talk, senior Tencent product manager He Zhichao discusses how to build a good data platform, sharing his thoughts and practices from his experience in data platform development.

He describes his personal journey from a software engineer to a data‑platform product manager, emphasizing the need for a blend of product instincts and deep big‑data technical knowledge.

The talk outlines five main parts: (1) thinking about the transition from technology to product, (2) deeply understanding the needs of data workers, (3) the practical implementation of Tencent’s Euler asset‑factory, (4) product‑manager methodology for data platforms, and (5) a Q&A session.

For the transition, he suggests evaluating personal traits, curiosity, responsibility, and opportunities, then gaining product‑thinking experience through simple prototyping and documentation.

Understanding data users involves recognizing the critical role of data in content‑apps, the lack of unified data‑asset tools, and the challenges of data quality, governance, and metadata management.

Euler’s asset‑factory aims to turn raw tables into well‑defined data assets, providing metadata collection, quality monitoring, and fast query via Presto, while addressing issues such as naming conventions, governance timing, and metadata completeness.

He proposes five solutions: embedding governance into production, integrating platform tools to avoid redundant metadata, improving root‑cause analysis of quality problems, adding evaluation metrics for data‑asset maturity, and applying software‑engineering practices (DataOps) to data engineering.

The product‑manager methodology stresses platform innovation, breaking down features into fine‑grained elements, focusing on core competencies, balancing standardization with efficiency, and driving user adoption through targeted incentives and continuous feedback.

The Q&A covers topics such as data model iteration, handling legacy quality issues, the impact of switching from Hive to Presto, governance of low‑usage tables, benefits of software‑engineering concepts in DataOps, and when a data platform is needed for mature domains.

Overall, the session provides a comprehensive view of data‑platform product development, combining technical depth with product‑management strategies to create scalable, maintainable, and user‑centric data services.