What matters for your decision
Use one fixed eligibility rule. If assignment is not random, call the result observational and record possible confounders before expanding the pilot.
Make the metric auditable
- Compare randomized eligible contacts or rotate queues at a predeclared time, with exceptions for urgent or sensitive work.
- Hold survey prompts and logging consistent before and after; capture verified resolution, escalation and complaints.
- Exclude unrelated operational changes or annotate them explicitly before interpreting impacts.