Choosing Test Strategies, Ensuring Test Stability, and Writing High‑Quality Tests

In traditional software development, testing is performed after coding, which often leads to late defect detection, higher repair costs, and increased risk. As the industry matures, teams are automating most tests to shift testing left, allowing early feedback before code submission.

However, extensive automated tests can raise maintenance overhead, lower test quality, and increase CI/CD resource consumption. Selecting suitable test strategies, ensuring test stability, and writing high‑quality tests are key to mitigating these issues.

1. Background

Testing traditionally occurs post‑implementation, causing delayed issue discovery and higher costs. Automation and test‑left practices aim to provide early visibility into software health.

2. Choosing an Appropriate Test Strategy

(a) Test Pyramid Classification

The test pyramid model suggests a layered approach: fast, small unit tests at the base, fewer integration tests in the middle, and a limited number of end‑to‑end tests at the top.

(b) Test Size Classification (Google)

Tests are categorized by size: small tests run in a single process without I/O, medium tests span multiple processes and may use network calls (typically localhost), and large tests run across machines for end‑to‑end validation.

(c) Evolution of Test Strategy

As projects grow, the proportion of test types shifts: low‑complexity systems may allocate 80% unit, 10% integration, 10% contract/end‑to‑end tests, while high‑complexity systems increase higher‑level tests and contract tests.

3. Ensuring Test Stability

(a) Local Testing

Local testing runs entirely on a developer’s machine without external dependencies, using in‑memory stubs or fakes for services, databases, and caches.

Two approaches are described:

In‑process: replace external services with in‑memory stubs and use in‑memory implementations for databases/Redis.

Cross‑process: minimize inter‑process communication and avoid network calls beyond localhost, using mock servers for services and in‑process fakes for storage.

(b) Test Doubles

Test doubles (Martin Fowler) replace real dependencies to improve speed and stability. Types include Dummy objects, Stubs, Spies, Mocks, and Fakes.

4. Writing High‑Quality Tests

(a) Test Case Design Methods

Common techniques include equivalence partitioning, boundary value analysis, error guessing, cause‑effect graphing, decision tables, orthogonal arrays, functional charts, and scenario‑based design.

(b) Naming Conventions

Test classes should be named ClassNameTest ; test methods should clearly describe the business scenario, using either snake_case or camelCase.

(c) Given‑When‑Then Pattern

Tests should follow the Given‑When‑Then structure to separate setup, action, and verification.

Conclusion

Selecting appropriate test strategies, ensuring test stability, and writing high‑quality tests are crucial for reducing maintenance cost, improving test quality, and accelerating feedback in software development. Future work includes exploring automated testing tools, AI‑assisted testing, and applying these practices across diverse project types.

References

Martin Fowler. https://martinfowler.com/

James Whittaker, Jason Arbon, Jeff Caro. Google Software Testing . People’s Posts & Telecommunications Press, 2013.