Can generated cases include the correct resolution, and can difficulty be varied?

**Can each case include its correct resolution?** A release can include a labelled root cause, an expected outcome, the evidence that supports the cause, and accepted alternative actions where more than one route is valid.

**Can difficulty be varied?** A buyer can ask for controls such as the number of systems involved, the number of decoy explanations, missing or partial links, ambiguous evidence, and how many steps separate the symptom from the cause. Scope which controls matter in a pilot.

**Can policies differ between cases?** Yes, that can be requested. The same 40.00 variance can be a write-off under one tolerance and an escalation under another, which tests whether a model reads the policy rather than assuming a default.

**Can the data include messy formats?** Cases can include inconsistent date formats, truncated references, abbreviations and rounding differences, along with free-text notes and emails that carry part of the evidence.

**Does generated mean random?** No. Relationships are constructed so that the cause can be determined from the evidence supplied. Cases meant to be unresolvable can be included and flagged in their metadata, so a model can be tested on saying that it cannot tell.

**Is any of it real business data?** No. It is generated, and grounding should be inspected in a sample.