An average response time can improve while a few customers wait much longer. It can also worsen because of one unusual case even when most requests are handled consistently. Choosing the measure requires understanding what question the team needs to answer.
Define a response first
Decide whether the clock stops at an automatic acknowledgment, the first human reply, or a useful answer. These events describe different experiences. If the business wants to understand when someone receives meaningful help, a receipt email should not count as completion.
Specify working hours, time zone, and any justified exclusions. Keep unresolved requests visible rather than measuring only the ones that already received a response. Otherwise the longest waits may disappear from the report.
Show the distribution
The median can describe the middle completed case, while a higher percentile can help reveal longer waits when the volume supports it. A simple distribution by time band may be easier for a small team to understand. Always include the number of cases.
Imagine a hypothetical business where most inquiries receive a same-day reply but a handful remain unassigned. A single average will not explain that ownership problem. A list of aging open requests alongside the summary provides a route to action.
Match the measure to a decision
Use typical response time to review the normal workflow, and the long-wait view to investigate exceptions. Avoid turning a metric into pressure for empty replies that stop the clock without helping the customer.
Review the definitions with the people answering inquiries. They should be able to recognize their work in the report and explain unusual cases.
Start by comparing the current average with a median, time-band view, and open-request count. The best combination is the one that makes both ordinary service and neglected exceptions visible, without adding more numbers than the team can use.

