Q18 of 38 · Test design
How do you choose between black-box techniques when you have limited time?
Short answer
Short answer: Match the technique to the input shape. Numeric ranges → BVA + EP. Boolean combinations → decision table or pairwise depending on count. State-dependent → state transition. Many independent inputs → pairwise. With less time, drop combinations of techniques and use one well rather than several poorly.
Detail
Black-box techniques aren't interchangeable — each fits a particular kind of input space, and using the wrong one wastes effort. The decision is "what's the input shape, what's the time budget, what risk are we managing?"
Match technique to input shape:
| Input shape | Best technique | Why |
|---|---|---|
| Numeric range | EP + BVA | Catches boundary off-by-ones cheaply |
| Few booleans (≤4) | Decision table or exhaustive | Reads as spec; small case count |
| Many booleans (5+) | Pairwise | Combinatorial explosion forces it |
| State-dependent behaviour | State transition | Bugs are at transitions |
| Use case-driven flows | Use case scenarios | Maps to acceptance criteria |
| Open-ended/unspecified | Exploratory | No tests can be designed in advance |
| Logical rules with constraints | Cause-effect graphing or decision table | Both model interactions |
Time-budget choices:
- 30 minutes: pick one technique that matches the dominant input shape. Better one technique applied well than three poorly.
- 2 hours: combine two — typically EP/BVA for the inputs plus pairwise (or a decision table) for the combinations.
- A day: layer all relevant techniques + an exploratory session at the end to catch what the formal methods missed.
Risk-based filtering:
- High-risk areas (payment, security, regulatory) → exhaustive coverage of the input space, even if costly. Pairwise alone isn't enough for compliance.
- Low-risk areas (admin tools, internal dashboards) → smoke + a handful of cases.
The senior move: explicitly choose not to apply a technique. "We're not running pairwise on this checkout because the combinations are already exhaustively covered by the decision table; pairwise would just duplicate." Articulating that decision is the seniority signal — it shows you're not reflexively layering techniques.