Which operational features make a routing benchmark worth building?

Use this checklist to identify operational features worth representing in a generated benchmark, choosing features whose omission would make an apparently correct assignment fail when the receiving team tries to act.

- Include requests submitted by representatives, where the agent must distinguish the person sending the message from the account holder whose service needs attention. - Include forwarded messages with obsolete subject lines, so the current request must be found in the message body rather than inferred from a heading. - Include internal shorthand whose meaning depends on the requesting department, with a supplied glossary that makes the interpretation accessible to the agent. - Include the same request arriving through different channels, where channel conventions affect available context but do not automatically determine the responsible team. - Include explicit customer preferences that conflict with responsibility rules, requiring the agent to distinguish a requested destination from an authorised destination. - Include requests describing a symptom shared by several business processes, with additional evidence that identifies which process is actually involved. - Include attachments that qualify the written request, ensuring the benchmark tests whether the agent reconciles evidence instead of treating the opening sentence as complete.