qc-test-design

QA Test Design

When to Use

• Writing test cases for a new feature
• Reviewing if existing test cases cover requirements adequately
• Designing tests for complex business logic with many conditions
• Planning exploratory testing sessions
• Optimizing test suite by reducing redundant tests

Core Jobs

1. Equivalence Partitioning (EP)

Group inputs into classes where the system behaves the same — test one value per class.

Example: Age field for insurance (must be 18-65)
Valid partitions:      [18-65] → test with 35
Invalid partitions:    [<18] → test with 10
                       [>65] → test with 70
                       [non-numeric] → test with "abc"
                       [empty] → test with ""

Result: 5 test cases instead of testing every possible age

2. Boundary Value Analysis (BVA)

Test at and around boundaries where defects cluster.

Example: Same age field [18-65]
Boundaries to test:
  Lower boundary: 17 (invalid), 18 (valid), 19 (valid)
  Upper boundary: 64 (valid), 65 (valid), 66 (invalid)
  Zero/null: 0, negative (-1), empty

Result: 8 focused tests covering the most defect-prone values

# BVA test pattern
@pytest.mark.parametrize("age,expected", [
    (17, False),   # below lower boundary
    (18, True),    # at lower boundary
    (19, True),    # above lower boundary
    (64, True),    # below upper boundary
    (65, True),    # at upper boundary
    (66, False),   # above upper boundary
    (0, False),    # zero
    (-1, False),   # negative
])
def test_age_validation(age, expected):
    assert validate_age(age) == expected

3. Decision Table Testing

For business logic with multiple conditions and combinations.

Feature: Loan approval
Conditions:
  C1: Credit score >= 700?
  C2: Income >= $50K?
  C3: Debt ratio <= 40%?

Decision table:
  C1 | C2 | C3 | Action
  Y  | Y  | Y  | APPROVE
  Y  | Y  | N  | REVIEW (high debt)
  Y  | N  | Y  | REVIEW (low income)
  Y  | N  | N  | DENY
  N  | Y  | Y  | REVIEW (low credit)
  N  | Y  | N  | DENY
  N  | N  | Y  | DENY
  N  | N  | N  | DENY

Result: 8 test cases covering all condition combinations

4. Pairwise Testing

When too many combinations exist — test every pair of inputs at least once.

Example: 3 parameters, 3 values each = 27 combinations
Pairwise reduces to ~9 tests while catching most pair-interaction defects

Tools: allpairs (Python), PictMaster, ACTS

from allpairspy import AllPairs

parameters = [
    ["Windows", "Mac", "Linux"],    # OS
    ["Chrome", "Firefox", "Edge"],  # Browser
    ["English", "French", "Spanish"],  # Language
]
for i, pairs in enumerate(AllPairs(parameters)):
    print(i, pairs)  # 9 pairs instead of 27

5. Exploratory Testing

Structured unscripted testing to discover unexpected defects.

Session-Based Test Management (SBTM):
  Charter: "Explore the checkout flow focusing on payment edge cases"
  Time box: 90 minutes
  Area: Payment processing

  Session notes structure:
  - What I tested: [actions taken]
  - Issues found: [bugs/observations]
  - Questions: [things to investigate further]
  - Coverage: [what I covered vs didn't]

Common heuristics (SFDIPOT):
  Structure: test the thing itself
  Function: what it does
  Data: what it processes
  Interface: interactions with other components
  Platform: environment, browser, OS
  Operations: how users actually use it
  Time: timing, concurrency, sequences

6. State Transition Testing

For workflows and state machines.

Example: Order status workflow
States: Pending → Confirmed → Shipped → Delivered | Cancelled

Test matrix:
  From\To    | Pending | Confirmed | Shipped | Delivered | Cancelled
  Pending    |    -    |   ✓ test  |    ✗    |    ✗      |   ✓ test
  Confirmed  |    ✗    |    -      | ✓ test  |    ✗      |   ✓ test
  Shipped    |    ✗    |    ✗      |    -    |  ✓ test   |    ✗

✓ = valid transition to test (happy + unhappy path)
✗ = invalid transition to test (should be rejected)

Key Concepts

• Equivalence Partitioning — divide inputs into groups with same behavior; test one per group
• Boundary Value Analysis — test at and just beyond boundaries; most defects live here
• Decision table — explicit mapping of condition combinations to expected outcomes
• Pairwise testing — test every pair of input combinations; finds most interaction defects with fewest tests
• Test oracle — the source of expected results (spec, domain expert, reference implementation)
• State transition — model system as states + transitions; test valid and invalid transitions
• Charter — exploratory testing mission statement defining scope, area, and time box

Checklist

• Equivalence partitions identified for each input field?
• Boundaries tested (at, just below, just above)?
• Decision table created for complex business logic (3+ conditions)?
• Invalid inputs tested (null, empty, overflow, special characters)?
• State transitions tested including invalid transitions?
• Exploratory sessions chartered and time-boxed?
• Test cases trace to requirements (coverage matrix)?

Key Outputs

• Test case suite covering all EP classes and BVA boundaries
• Decision table for complex business logic
• Exploratory testing charter and session notes
• Requirements coverage matrix (which tests cover which requirements)

Output Format

• 🔴 Critical — only happy path tested (no boundary/invalid cases), no negative test cases, copy-paste test cases that test the same thing
• 🟡 Warning — random input selection instead of systematic techniques, no exploratory session for high-risk features
• 🟢 Suggestion — apply pairwise for large parameter spaces, create state transition diagram before writing tests, use heuristics checklist for exploratory sessions

Anti-Patterns

• Testing only the obvious happy path (bugs live at boundaries and in combinations)
• Writing 100 test cases that all test the same equivalence class (wasteful)
• Skipping exploratory testing entirely (scripted tests only find known unknowns)
• Decision tables with missing condition combinations (coverage gaps)

Integration

• test-driven-development — TDD drives test design; use these techniques to define the tests
• qc-automation — automate test cases designed here using the framework patterns
• test-strategy — these techniques feed into the overall test strategy