Building and Managing a Metric System: Methodology, Lifecycle, and Productization at Didi

Metrics are systematic collections of related indicators that provide a global view of business health; they consist of both result‑type and process‑type metrics and are organized along dimensions such as qualitative and quantitative attributes.

The metric lifecycle includes definition, production, consumption, and deprecation, requiring continuous operation, quality assurance, and data‑driven governance.

Metric selection can follow the OSM (Objective‑Strategy‑Measurement) model or the classic AARRR (Acquisition‑Activation‑Retention‑Revenue‑Referral) pirate model, enabling systematic coverage of growth stages.

Scenario‑based construction uses the "person‑product‑scene" abstraction (e.g., Didi's ride‑hailing: users, services, and channels) to derive relevant indicators and dimensions such as city, time, and user tags.

Management pain points are identified from business, technical, and product perspectives, highlighting issues like unclear analysis scenarios, inconsistent metric definitions, duplicated data pipelines, and lack of unified consumption interfaces.

Management goals focus on technical unification of metric naming, calculation, and source; business alignment of data outputs and scenario coverage; and productization of metric management tools to support decision‑making, analysis, and operations.

The architecture defines data domains, business processes, time cycles, modifier types, atomic metrics, derived metrics, and dimensions, with clear classifications for atomic, derived, and composite metrics.

Metadata management separates dimension management (basic and technical information) from metric management (basic, technical, and derivation information), enabling traceability and governance.

The end‑to‑end workflow includes modeling, development, and deployment phases, supported by a data‑warehouse layer (DWM) that stores core metric data.

Metric system productization is realized through a dictionary tool that standardizes metric definitions, ensures uniqueness, and provides a unified API for downstream consumption.

Overall, the article presents Didi's metric system methodology, tooling, and roadmap for scaling data‑driven decision making across the organization.