How Huolala Built a High‑Impact Metric Library to Power Data‑Driven Decisions

Background

Huolala is a technology company focused on the freight industry, providing online freight services such as intra‑city and inter‑city transport, enterprise logistics, and less‑than‑truckload shipments. The company processes massive amounts of business data daily to understand user needs, optimize services, and improve operational efficiency.

Goals

The self‑built metric library aims to:

Data‑driven decision making : Leverage data as a core basis for business decisions.

Improve operational efficiency : Monitor business performance in real time, detect issues promptly, and adjust quickly.

Optimize services : Analyze metric data to understand user behavior and enhance satisfaction.

Support business development : Provide insights for strategic planning and execution.

System Overview

The metric library is a systematic, structured platform for managing metric metadata. It stores metric names, meanings, calculation formulas, data sources, classifications (e.g., financial, marketing, production), historical and current data, and analysis reports. By analyzing and comparing metrics, businesses can assess status, identify problems, and adjust strategies.

Metric Definition

Metrics consist of several components:

Atomic metrics : Fundamental data points that directly reflect basic business conditions and trends.

Modifiers : Qualifiers such as time range, region, or business scope that add specificity.

Statistical periods : Time windows (day, week, month, quarter, year, or custom cycles) used for calculation and analysis.

Metric metadata : Descriptions of meaning, purpose, and calculation methods that aid interpretation and usage.

Application Scenarios

AB‑Test Experiment Report

Project background : By establishing a robust data capability, Huolala linked the metric library to automate AB experiment report generation, greatly enhancing efficiency and accuracy.

Project benefits : The system now automates about 80% of freight experiment reports, covering roughly 45% of conversion‑rate usage and 25% of deep‑usage conversion, providing comprehensive data support for decision‑making.

Technical chain :

Experiment data pipeline (data warehouse side + engineering side)

Engineering side data integration pipeline

Metric dashboard generation

Metric Encyclopedia

The library integrates with Feishu dictionary, allowing quick lookup of metric definitions, scopes, and related information directly within Feishu documents and chats.

Multi‑Factor Attribution

Supports data‑link diagnosis, correlation diagnosis, and DuPont analysis to trace the impact of data across different metric layers.

Achievements and Planning

Achievements

High coverage : Over 80% of core metrics are covered.

Strong universality : More than 10 dimensions and 200+ metrics serve common analysis scenarios; core reports have integrated the metric system.

Broad applicability : The library powers data analysis, metric alerts, anomaly attribution, and other use cases.

Future Plans

Efficiency and stability : Build a high‑performance system that remains smooth under large data loads.

Data security : Implement strict data protection and encryption measures.

User‑friendly design : Offer an intuitive interface and detailed guides for easy onboarding.

Rich functionality : Expand capabilities to include advanced visualization, forecasting, and more.