How Heartbeat Game Built a Cloud‑Native Big Data Platform with Alibaba DataWorks

Overview

Heartbeat Game, founded in 2003, operates dozens of games worldwide with over 50 million monthly active users. To support AI, risk‑control, and analytics scenarios, the company built a cloud‑native big data platform on Alibaba Cloud using DataWorks.

Platform Architecture

The platform follows a three‑layer architecture: data ingestion, storage, and application. Data is collected via SLS and Kafka, cleaned with Flink, and stored in OSS and Hologres. Batch processing is handled by MaxCompute with DataWorks, while real‑time analysis combines Flink and multiple OLAP tools.

Universal Game Data Model (GS‑LBDM)

A logical data model (LDM) based on 3NF organizes game data by themes such as player, product, SDK, event, and release. The model supports diverse analytical applications, including AI, risk control, and data analysis, and is built on DataWorks’ intelligent data modeling capabilities.

Logical Data Model Details

The GS‑LBDM defines six core domains: Player, Product, Event, Release, Studio, and Region. Each domain is modeled independently yet linked, enabling seamless data integration across games.

Player Domain Example

The player domain follows the AARRR framework (Acquisition, Activation, Retention, Revenue, Referral). Metrics such as DAU, MAU, ARPU, CAC, and LTV are captured, with dimensions like time, region, version, and channel to evaluate marketing effectiveness.

Implementation with DataWorks

DataWorks provides four main modules: data warehouse planning, data standards, data metrics, and dimensional modeling. Planning defines layers (ODS, CDM, ADS) and naming conventions. Standards enforce consistent field definitions. Metrics enable bulk creation of derived indicators. Dimensional modeling supports both forward and reverse engineering, allowing reuse of existing physical tables.

Benefits of GS‑LBDM

Extensibility : Adapts to business changes with minimal model modifications.

Flexibility : Third‑normal‑form design is platform‑agnostic, reducing redundancy.

Reusability : Consistent data definitions across multiple games facilitate cross‑game analytics.

Implementation Ease : Rich documentation and sample tables accelerate deployment.

Conclusion

The GS‑LBDM offers a scalable, flexible, and application‑neutral data foundation that empowers game developers to quickly derive insights, improve player experience, and maximize revenue.