How Xianyu Built a Scalable Test Data Generation Platform for Faster Testing

Background

As Xianyu’s business grew rapidly, the variety of product types, transaction templates, and interactive features increased, causing test data creation to consume large amounts of tester time. The main problems identified were:

High labor cost : Hundreds of product and order state combinations required lengthy, labor‑intensive data construction.

High entry barrier : Data creation is tightly coupled with account types and user groups, demanding extra effort for test acceptance and cross‑team collaboration.

Lack of data support for test tools : Automated and performance testing need rich data types as drivers.

To address these issues, Xianyu’s testing team built a fast‑onboarding test‑data generation solution usable on PC, the Xianyu app, and DingTalk.

Solution Design

The platform’s overall architecture connects multiple business lines and supports data construction for products, orders, and promotions, while also feeding data to automation tools. It aims to let partners obtain data conveniently during product acceptance.

A second diagram shows the internal design focused on extensibility and ease of use: a plug‑in‑style configuration lets users build custom data‑generation scenarios, and the platform integrates with various front‑ends for agile data creation.

Dynamic Configuration Support

To simplify onboarding for different businesses and provide data for CI/CD, automated inspection, and API testing, the platform adopts a plug‑in data‑source model. The process is:

Configure business metadata on the platform.

Create business‑specific templates based on the metadata.

Share the same configuration across PC, the Xianyu app, and the DingTalk bot.

Multi‑Endpoint Access

The platform offers three entry points, each tailored to a user group:

Xianyu App

The app can automatically capture device environment information. Using a JS Bridge, the current user’s login state is obtained, enabling one‑click product publishing and immediate retrieval of the product schema for rapid verification.

DingTalk Bot

The bot provides a convenient, generic, and lightweight way to generate data within daily workflows. For example, the “inspection treasure” business (C2S2C) uses the bot to push orders, eliminating the need to contact developers or service providers and allowing error feedback directly in the chat group.

PC Workbench

The PC console excels at management tasks. Users can configure and customize product publishing and order templates, clone templates, and perform batch data generation, following the dynamic configuration steps described earlier.

Improving Data Coverage

The platform currently covers three major business lines—products, transactions, and marketing discounts—and can construct data for product details, order fulfillment, and promotional offers. The target data types are illustrated in the following diagram.

All major product types are now fully covered; transaction data supports various C2C order states, with additional order types being integrated continuously.

Results and Outlook

Before the platform, testing, product acceptance, and cross‑team collaboration required extensive manual data creation. After deployment, the platform’s configurable and extensible design enables self‑service data retrieval, dramatically boosting efficiency. Product publishing speed improved from minutes to seconds, and a full order‑fulfillment cycle dropped from about an hour to a few minutes.

Since its launch, the platform has onboarded product, transaction, and discount lines, covering over 20 core product types, publishing more than 60,000 products and generating over 100 orders. It also supported data preparation for product compliance and detail‑page upgrades, further enhancing test efficiency.

Continue expanding data coverage to include order diagnostics, product diagnostics, and user‑asset data.

Introduce custom plug‑in extensions to lower integration effort.

Enrich the “test‑Ding integration” interaction model so that a single chat group can satisfy all data‑fetching needs.

The ultimate goal is to empower business testing with rapid data generation, promote left‑shift testing, and free Xianyu engineers from repetitive manual work.