How HuoLala Built a 0‑to‑1 Stability Metric System and Cut Faults by 78%

Background: HuoLala operates a large on‑demand logistics platform covering 360 Chinese cities with over 950 000 daily active users and 680 000 drivers. Such scale makes technical stability essential.

Why Build Stability Metrics?

Metrics turn vague feelings of stability into quantifiable performance results, provide clear targets, and drive continuous improvement of the stability system.

Initial Pain Points

Metrics were scattered and not linked to business goals.

Definitions were ambiguous, causing inconsistent interpretation.

Data collection was labor‑intensive and error‑prone.

Stability measurement was treated as a one‑off activity rather than a systematic process.

Principles for a Sustainable Metric System

The system must be a long‑term effort, goal‑oriented, and supported by platform tools that automate data collection and analysis. An operating mechanism should regularly observe, diagnose, and remediate issues.

Core Process of Building the Metric System

Analyze Current Pain Points – Identify gaps in existing metrics and data pipelines.

Define Clear Principles – Ensure metrics are goal‑driven, measurable, and maintainable.

Execute Core Tasks

Define Metrics : Start from high‑level stability KPIs (e.g., fault count, SLA) and decompose them into hierarchical indicators such as continuous no‑fault days, fault severity levels, and stage‑wise response times.

Collect Data : Replace manual Excel tracking with a dedicated platform that automates data ingestion, ensures accuracy, and supports historical analysis.

Operate the System : Run the metrics from top‑down (executive reporting) and bottom‑up (team‑level KPI breakdown), and share insights internally to foster a stability‑first culture.

Metric Definition Example

From the stability goal, HuoLala set a fault‑count target and defined fault severity tiers based on impacted services. Additional milestones such as "continuous 180 days without a fault" were introduced to motivate the team.

Data Platform

A global stability dashboard aggregates core metrics (fault count, SLA trends, no‑fault days) and domain‑specific views (emergency response, change control, post‑mortem efficiency). The platform supports dimension drilling (by month, department) and enables proactive alerting.

Cultural Adoption

An internal "IP" (a virtual NOC persona) was created to publish regular stability summaries, highlight achievements, and educate teams. Consistent communication turned stability metrics into a shared language and habit.

Results

After more than two years of effort, HuoLala reduced fault count by 78% year‑over‑year and improved SLA from three nines to four nines. The reduction translated into roughly eight fewer fault‑hours per year for business users.

Iteration & Future Plans

The metric system is not a one‑off project; it requires continuous iteration. Upcoming focus areas include deeper data mining to uncover root causes, smarter anomaly‑detection for early warnings, and richer platform capabilities for flexible metric composition and multidimensional analysis.

Conclusion

Stability metric measurement is analogous to a health check‑up: it requires a set of well‑defined examinations, regular reporting, and continuous follow‑up. By establishing a systematic, data‑driven metric framework, HuoLala created a virtuous cycle that improves reliability, reduces downtime, and aligns the whole organization around a common stability goal.