How QianNiu Cut Plugin Issues by 50%: Open Architecture & Monitoring Secrets

QianNiu is Alibaba’s multi‑endpoint open work platform that serves millions of merchants on both mobile and desktop, providing store management, customer service, and information messaging.

Open Architecture and Issues

The platform defines a set of open nodes (plugins) with standard protocols, allowing third‑party ISVs to develop and integrate services. A third‑party plugin follows four stages: ISV development, marketplace listing, merchant purchase, and usage. Rules guide free trials and upgrades, but the openness also introduces merchant experience problems.

Open Nodes

Problem Characteristics

The open experience issues are highly scattered and have complex causes. Numerous tools from various ISVs, differing maintenance levels, and a heterogeneous tech stack (early H5, mid‑stage QAP, and mini‑programs) result in a long plugin startup chain involving more than seven technical steps, making the system vulnerable to network and service fluctuations.

Core Experience Problems

Long plugin startup chain affecting launch success.

Permission restrictions for sub‑accounts causing blocking errors.

ISV or second‑party logic issues and insufficient online problem perception.

Overall Solution

Optimize plugin startup chain to improve fault tolerance and raise success rate above 99.7%.

Build a closed‑loop permission request process to enhance approval efficiency for sub‑accounts.

Establish a data‑driven measurement system to guide business optimization.

Plugin Startup Chain Optimization

Protocol routing: parse and route according to defined open node protocols.

Plugin meta‑information lookup: locate the target plugin’s metadata from the server‑issued list.

Permission verification: check sub‑account permissions for the plugin.

Commercial guarantee: ensure free‑version subscription for new or expired users.

Pre‑auth QAP: trigger explicit user authorization for third‑party plugins.

Container routing and rendering: assemble business parameters and hand over to the appropriate container.

Permission Request Chain

The mobile side adds a permission request flow: client APIs are exposed to second‑party ISVs to trigger application prompts; automatic detection of permission shortages in third‑party links triggers request prompts; approval messages are sent to the main account for review.

Data Measurement System

A comprehensive metric system was built, covering technical indicators (interface success rate, bridge API success, JS error rate) and perception metrics such as white‑screen rate. This system drives online issue detection, rollback, and fixes.

White‑Screen Detection

White‑screen is defined as a period where page elements fail to render, resulting in a blank or error page. Detection strategy counts effective elements (text, images) via injected JS in WebView, filters noise scenarios (auto‑redirects, authorization screens, special component pages), and excludes known white‑list phrases.

Case Studies

Second‑party business experienced white‑screen on 4G; timely alerts reduced the issue and prevented complaint spikes.

Third‑party plugin triggered API rate limiting; after product‑level adjustments, success rate improved significantly.

Complaint SOP

When users submit feedback, the system records the plugin usage path and attaches the plugin appkey to the complaint, guiding users to select the relevant plugin. This increased plugin‑related complaint proportion from 11.95% to 95%.

Open Experience Dashboard

By integrating complaints, technical metrics, and perception data, an open‑experience dashboard provides a holistic view of plugin quality, enabling traceable, measurable, and actionable improvements.

Conclusion

QianNiu enhanced plugin startup fault tolerance, built a permission request chain for sub‑accounts, and established a data‑driven measurement system. The open‑experience dashboard now offers a comprehensive view of plugin performance, allowing rapid issue detection and targeted optimization, ultimately improving merchant experience on the platform.