Xiaomi Metric System Construction and Management Best Practices

Overview : The article shares Xiaomi's best practices for building and managing a metric system, organized into five parts: system overview, business pain points and solutions, construction methods, Xiaomi's own practices, and summary with outlook.

1. Metric System Overview : A metric system consists of two aspects—metrics (quantitative measures of business units) and dimensions (different perspectives that together form a complete system). The core goal is to create appropriate metrics and dimensions to support business analysis.

2. Business Pain Points and Solutions : Three main challenges are identified from business, technical, and product perspectives: inconsistent metric definitions, fragmented data sources, and lack of end‑to‑end visibility. The proposed solution involves unifying metric definitions, consolidating data pipelines, and providing a single platform for metric consumption.

3. Construction Methods : Xiaomi adopts two key methodologies:

OSM Model (Object‑Strategy‑Measure): defines business goals, strategies, and measures to align metrics with strategic objectives.

MECE Principle (Mutually Exclusive, Collectively Exhaustive): ensures metric dimensions are non‑overlapping and comprehensive, using techniques such as binary division, process mapping, element analysis, formulaic classification, and matrix methods.

4. Xiaomi's Best Practices : The company integrates the OSM model and MECE principle to design a unified data warehouse, establishing a consistent data dictionary, standardized metric definitions, and a semantic layer that powers self‑service BI dashboards.

5. Implementation Path :

Model Design: Apply MECE to partition data domains and build core models using dimensional modeling.

Data Warehouse Construction: Follow unified standards to create a stable, secure corporate data warehouse.

Metric Management: Maintain a unified data dictionary to ensure consistent metric definitions.

Data Application: Enable business users to discover metrics via a data encyclopedia, generate BI dashboards, and share insights.

Evaluation Principles : Seven principles guide model design—high cohesion & low coupling, separation of core and extension models, centralized common logic, cost‑performance balance, data rollback capability, consistency, and clear naming conventions.

Future Outlook : Xiaomi's metric system links data development, metric management, and analysis, achieving full‑link visibility. The methodology and tools have been rolled out across 55 data domains and over 520 core atomic metrics, with ongoing iteration and promotion.

Overall, the article demonstrates how a well‑structured metric system and data platform can reduce data usage costs, improve decision speed, and align business units with corporate goals.