Building an Attribution System for NetEase Cloud Music Data Warehouse: Challenges and Solutions

The presentation introduces the attribution system built for NetEase Cloud Music, outlining three main sections: the problems encountered, the proposed solutions, and future planning.

Problems : The existing data pipeline integrates client logs, server logs, algorithm tags, and business data into a dimension‑DWD‑DWS architecture, but attribution suffers from high latency due to massive log sorting, unclear attribution results, and limited extensibility for new business scenarios.

Attribution Background : User behavior attribution aims to trace the cause of conversion actions (play, like, comment, purchase) across complex app flows, focusing on the last touchpoint that adds content to the playback list rather than the direct UI navigation.

Previous Implementation Issues : The prior ETL‑based approach required costly sorting, produced ambiguous results, and could not handle cross‑day or multi‑level attribution, leading to stability, accuracy, and scalability challenges.

Solution Overview :

1. Event‑tracking framework upgrade : Jointly developed by the data warehouse and front‑end teams, introducing a standardized SDK that records object IDs, positions, content types, and distribution strategies, eliminating large‑scale log joins.

2. Attribution model upgrade : Defines three attribution parameters (PS refer, Multi refer, add refer) and distinguishes between accompanying and non‑accompanying states to capture richer context for playback, comment, and share events.

3. Management platform : Provides unified configuration for requirements, objects, and events, generating dimension tables for downstream attribution analyses.

The upgraded model supports multi‑dimensional analysis across page positions, media content, business scenarios, and traffic classifications, enabling more precise and extensible attribution.

Future Planning : Expand coverage of attribution tracking across client types, introduce first‑touch attribution for entry analysis, standardize SPM definitions across business lines, and consolidate all configuration rules within the management platform to further boost efficiency and reliability.

Overall, the new system improves attribution data stability, accuracy, and extensibility, supporting offline, near‑real‑time, and real‑time analytics for NetEase Cloud Music.