Write the expectation before running the system
List rules in decision order with their inputs, recipients and exceptions. Prepare marked fictional cases covering boundaries: correct category, missing category and a value from an older form version. Set expected ownership for each row before testing.
In a fictional trial, project requests go to projects, existing customer support goes to support, and unknown values go to triage. A case appearing in a shared inbox is not a pass if the agreed team assignment is missing.
Separate queuing from personal acceptance
Agree evidence for each step. A technical acknowledgement can establish receipt of a record. Queue state establishes placement. An owner field or acceptance event can establish responsibility under your model. Do not substitute one evidence point for all the steps.
Also test an absent or disabled recipient with authorised accounts. Follow the chosen fallback. Where the system has no absence information, the trial must not assume the integration knows it.
| Test case | Expected result | Keep |
|---|---|---|
| Known category | Correct queue or agreed owner | Case reference and rule version |
| Missing decision value | Chosen triage route | Exception rule used |
| Disabled recipient | Fallback or visible stop | Error state and responsible owner |
| Two matching rules | Documented priority | Winning rule and reason |
| Same test delivered again | No unintended additional owner | Relationship to original case |
Test collisions and rule changes
Include a case matching both a general and a specific rule. Check agreed priority and that a change does not move historical cases unintentionally. Leaving older cases in place may be correct, but the decision should be explicit.
On repeated delivery, distinguish a newly created case from an update to an existing record. Compare the stable test reference and ownership history. Many log activities do not establish many new enquiries or correct assignment.
Decide who detects an unowned case
A case without responsibility needs a defined detection routine. Specify what counts as missing ownership, who reviews the list and how exceptions escalate. Set the threshold around real working practices rather than an invented universal response promise.
Finish with observed result, deviation and retest after correction. Remove fictional cases under your plan. Repeat important recipient checks when people, categories or form values change. The trial records the controlled outcome, not every future case.