LEARN. PRACTICE. GROW.Concept and mechanism
A decision table makes actions for condition combinations explicit. Two independent Boolean conditions produce four combinations before constraints are considered. If one condition no longer affects the action within a subset, rules may be merged, but behavior must remain correct for every represented combination. Counting covered combinations helps find gaps; a correct sum does not prove equivalence when overlap and omission coexist. Retain a justification for infeasible rules. For a complex authorization rule, discuss the table with the behavior owner before automating test generation. An incorrect model can generate many tests consistent with the same mistake.
Guided application
When calculating an exact outcome is difficult, a metamorphic relation can compare related executions. In an exporter where order has no meaning and totals use integer cents without sequence-dependent fees, reordering the same transactions should preserve the total. If the relation fails, at least one premise or result needs investigation; the relation alone does not identify which execution is wrong. Complement directed techniques with exploration guided by a mission: investigate recovery after interruption within a defined time and environment, recording observations and questions. A checklist supports memory and experience but should evolve with discovered defects. Neither rule count, session duration, nor automatic generation replaces evidence evaluation.
The same coverage sum can hide one repeated combination and another missing.
Common pitfalls
Model treated as truth; relation without premises; exploration without records.
Related topics: Analysis and evidence in the test process · Priority and residual risk · Domains, boundaries, and combinations
Choose useful models and investigate their limits.
Reference: Satisfice exploratory testing · CTAL-TA v4.0 (2025)