Front-end UI Automation Testing: Challenges, Solutions, and Practices

This article shares the author’s experience of building and maintaining front‑end UI automation across desktop, web, and mobile platforms. It begins with a humorous collection of test‑engineer jokes that illustrate the impossibility of testing every scenario, setting the stage for the importance of continuous‑integration automation.

The early expectations of automation—reducing error rates, improving coverage, and speeding feedback—have given way to practical concerns such as human cost, maintenance effort, and diminishing returns. The author outlines the evolution of the automation stack, from simple Selenium scripts to integrated platforms like QTA, Wetest, and custom SDKs, and discusses the trade‑offs of using third‑party frameworks (AutoIt, Puppeteer, Appium, etc.) versus building in‑house solutions.

Key problems identified include:

Business complexity and a large number of test cases, making end‑to‑end tests costly and fragile.

Frequent merges that break large test suites, causing massive regressions.

Low efficiency and high cost due to manual test‑case understanding, script maintenance, long execution times, and difficult failure analysis.

To address these issues, the article proposes a three‑pronged approach:

Process planning and result optimization: unify frameworks horizontally across products and vertically across platforms, share common code for login, initialization, and driver upgrades, and standardize test‑case structure (description, implementation, API layer).

Team collaboration and case sharing: create shared test tasks, encourage cross‑team ownership, and integrate testing into the development pipeline (e.g., trigger on merge requests).

Report improvement: redesign reports for clarity, add automatic error classification, enable one‑click bug filing, and provide coverage metrics.

Practical techniques highlighted include:

Using platform_type = config.Platform_Type to select the appropriate test base class for web, Windows, or macOS.

Encapsulating common UI actions (e.g., bold, font change) in reusable methods like self.startStep('设置加粗') followed by logic.set_bold() .

Replacing brittle XPATH selectors with image‑template matching or JS‑based property extraction.

Leveraging ADB broadcasts for Android UI actions (e.g., adb shell am broadcast -a com.tencent.xxx --es action click_accept ).

Results after several months of optimization include:

Core business coverage exceeding 70% with a 93%+ pass rate on daily change‑set tests.

Daily automation runs over 150 times, a test‑case library of 6,000+ cases, and a reduction in manual testing effort.

Improved defect discovery (140+ bugs, 60+ critical) and faster feedback loops for developers.

The author concludes that automation must stay tightly coupled with business needs, continuously evolve, and be measured with clear metrics to remain sustainable. Future work will focus on AI‑assisted test generation, further reducing flakiness, and expanding coverage to performance and stability testing.