Testing Models in QA

1. Conventional Testing

• Definition:
  Conventional (or traditional) testing means designing test cases manually based on functional requirements, user stories, or test scenarios. It does not follow a specific statistical or combinatorial approach.

• Example:
  For a login system with 3 browsers (Chrome, Firefox, Edge), 2 OS (Windows, Linux), and 2 languages (EN, FR), a tester might manually select test cases like:

  • Chrome + Windows + EN

  • Firefox + Linux + FR
    (but may miss some combinations).

• Industry Suitability:

  • Small applications with limited input combinations.

  • Domains where human intuition and exploratory testing are crucial (e.g., UI/UX testing, small web apps).

---

2. Pairwise Testing (All-Pairs Testing)

• Definition:
  A black-box test design technique where test cases are chosen such that every possible pair of input parameter values is covered at least once.

  • "All-Pairs Testing" is simply another name for Pairwise Testing.

• Example:
  Parameters: Browser (Chrome, Firefox, Edge), OS (Windows, Linux), Language (EN, FR).
  Instead of 12 exhaustive combinations, Pairwise Testing may reduce this to 6–7 test cases, ensuring every pair (e.g., Chrome+Linux, Firefox+Windows) appears at least once.

• Industry Suitability:

  • E-commerce (testing checkout workflows across browsers + payment methods).

  • Mobile app testing (device + OS version pairs).

  • Embedded systems where two parameters’ interactions are most critical.

---

3. Combinatorial Testing

• Definition:
  A generalization of pairwise testing. It systematically covers input combinations with a defined strength:

  • 2-way (pairwise)

  • 3-way

  • … up to n-way (all combinations).

• Example:
  Using the same login system (3 × 2 × 2 = 12 combos):

  • 2-way testing → ensures every pair of inputs appears. (~6 cases)

  • 3-way testing → ensures every triple appears. (~12 cases, i.e., exhaustive here).

• Industry Suitability:

  • Aerospace, automotive, healthcare → safety-critical systems where multi-parameter interactions matter.

  • Telecom → protocol testing where 3–4-way interactions often matter.

---

4. Orthogonal Array Testing (OAT)

• Definition:
  A statistical design of experiments (DoE) technique using orthogonal arrays (balanced matrices).
  Ensures each parameter value is tested equally and in combination with others, but in a mathematically structured way.

• Example:
  Using an L4 orthogonal array for the login system may reduce 12 cases to 4 balanced test cases like:

  Test Case	Browser	OS	Language
  1	Chrome	Windows	EN
  2	Chrome	Linux	FR
  3	Firefox	Windows	FR
  4	Edge	Linux	EN

  This ensures uniform coverage of each value without testing exhaustively.

• Industry Suitability:

  • Manufacturing, Six Sigma, Automotive → originally from industrial quality control.

  • Telecom, Semiconductor, Embedded Systems → where statistical sampling is preferred.

  • Software QA → when input parameters are numerous but need balanced coverage.

---

5. All-Pairs Testing

• Already covered: this is another term for Pairwise Testing.

• Used interchangeably in software QA.

---

Comparison Table

Aspect	Conventional Testing	Pairwise / All-Pairs	Combinatorial	Orthogonal Array Testing (OAT)
Approach	Manual, ad-hoc	Covers all 2-way pairs	Covers n-way combos	Statistical, balanced sampling
Test Case Count	Unpredictable (can miss)	Moderate (fewer than exhaustive)	Higher (depends on n)	Low (optimized array-based)
Coverage	Depends on tester	All pairs covered	All interactions (up to n)	Balanced, uniform coverage
Strengths	Flexible, intuitive	High defect detection with fewer cases	Detects higher-order interaction defects	Optimized, mathematically proven balance
Weaknesses	Risk of missing defects	Misses higher-order (3+ way) bugs	More test cases as n increases	Harder to design arrays for large inputs
Best For	Small/simple apps	Web, e-commerce, mobile	Safety-critical, telecom, complex systems	Manufacturing, telecom, software sampling

---

Which is Best in QA?

• For small projects → Conventional Testing (quick, intuitive).

• For web & mobile apps → Pairwise/All-Pairs (great balance of effort & defect detection).

• For safety-critical/complex systems → Combinatorial (3-way or higher).

• For manufacturing/telecom/optimized sampling → Orthogonal Array Testing.

• 👉 Overall Best (most practical in QA/software): Pairwise (All-Pairs) Testing — because it offers the best balance between coverage, defect detection, and efficiency for most software applications.