Building a Data Metric System for NetEase Media Using OSM and AARRR Models

What is a metric system? A metric system is an organized collection of inter‑related statistical indicators that together reflect the overall quantitative characteristics of a business, consisting of metrics and dimensions.

Why build a metric system? In the era of fine‑grained operations, accurate and complete data metrics are crucial for product growth, business evaluation, and aligning team goals.

How to build the metric system?

1. North Star Metric : Identify the single most important metric that guides the company’s direction, following the SMART principle.

2. OSM Model (Objective‑Strategy‑Measurement): A three‑step framework that breaks down high‑level objectives into actionable, measurable strategies.

3. AARRR Model (Acquisition, Activation, Retention, Revenue, Referral): A set of five key stages for evaluating startup metrics, often combined with other models.

NetEase Media Data Warehouse Architecture

The platform follows a layered design: ODS/ODM (raw data), DWD/EDM (detail layer), DWS/GDM (aggregation layer), ADM (application layer), and DIM/DDM (dimension layer). Table naming conventions encode business line, source, layer, and update frequency, e.g., odm_{business}_{source}_{db}_{table}_incr_day .

Implementation Steps

5.1 Preparation – configure subject domains, layer settings, and dictionary collections in the model design center.

5.2 Dimension Construction – search and create dimensions via the UI.

5.3 Metric Construction – define metrics, add modifiers, and set calculation logic.

5.4 Table Construction – follow naming rules, define fields, and link dimensions and metrics.

5.5 Benefits – the metric system provides clear lineage, easy query of metric usage, and consistent metric definitions across tables.

Future Outlook

The article concludes that NetEase’s big‑data platform has become the core tool for data warehousing, supporting dimensions, metrics, and table design, and encourages broader participation from all roles in the data ecosystem.