- Observe
- Identify legal values, behavioral partitions, boundaries, transitions, X or Z relevance, constraints, and the exact sampling context.
- Decide
- Choose which values belong together, which need independent bins, and when arrayed, transition, wildcard, default, ignored, or illegal bins add meaning.
- Failure
- Automatic partitions can produce a reassuring metric with no requirement behind it, while excessive splitting creates noise and slows closure.
- Prove
- Trace bins to plan items, exercise representative boundaries, and audit expected versus observed counts at the intended sampling point.