Metric Management and Intelligent Insight Practices at MuTong Technology

The presentation introduces MuTong Technology's practice in metric management and intelligent insight, outlining five major parts: metric management, production‑consumption, naming conventions, definition & governance, and a Q&A session.

Metric Management Pain Points : difficulty finding data, low efficiency, poor quality, and low ROI caused by fragmented data sources, inconsistent definitions, and high AWS compute costs.

Strategy & Functional Layout : centralize product and warehouse metrics, establish a two‑level indicator system, standardize business and technical definitions, embed policies into the workflow, and provide self‑service analysis, smart exploration, and knowledge‑base tools.

Metric Naming : use a "business process + entity + measure" pattern for atomic metrics, derive extended metrics by adjusting periods or adding modifiers, and separate atomic, derived, and composite metrics across DM, ADS, and OLAP layers.

Metric Definition & Governance : unify definitions, bind data sets, enforce naming and term standards, implement checklist‑driven intake, conduct smoke/white‑box/black‑box testing, and monitor quality through pre‑, mid‑, and post‑stage controls, including ROI‑based deprecation.

Demand Process : integrate metric considerations early in requirement analysis, clarify dimensions and data sources, align product and analysis teams, and standardize documentation to reduce iteration cycles.

Metric Production & Consumption : build standardized domain models, bind datasets for real‑time and analytical use, define a DSL for query requests, and optimize execution plans with model matching, OneSQL stitching, and materialized view acceleration.

Intelligent Insight : identify deterministic and potential issues, use DAU case studies to illustrate root‑cause analysis via dimension and metric decomposition, and generate automated insight reports with visual cues.

Metric Governance : establish quality assurance processes, monitor hotness and ROI, label business and technical components for cost allocation, and continuously refine cost‑optimization strategies, achieving significant AWS cost reductions.

Q&A : address legacy metric governance, data classification and grading, and the use of AI/ML techniques for automated data categorization.

The session concludes with a summary of the shared practices and thanks the audience.