Choose the event and the clock
Decide whether the metric measures first automated acknowledgment, first useful answer or first human reply. Report those separately. Start the clock at the agreed customer event and state whether you count elapsed time or staffed time. US and Saudi schedules can differ, so store a consistent timestamp and apply the documented coverage calendar when calculating a business-hours metric.
Inspect the distribution and the excluded cases
A mean can hide a long tail of customers waiting much longer. Report a median and a slower percentile when the sample supports it, with the conversation count. Explain how spam, reopened cases and messages arriving outside coverage are handled. Compare similar intent and language groups so a new mix of easy questions is not mistaken for a faster workflow.
Fictional worked example
Fictional conversation: received at 9:00, automatic receipt at 9:01, useful answer at 9:12 and human reply at 9:20. Those are three different measurements: 1, 12 and 20 minutes of elapsed time. Reporting only the one-minute acknowledgment as a service improvement would obscure how long the customer waited for help.
Action checklist
- Define the first response event and elapsed or staffed clock.
- Keep acknowledgment, useful answer and human reply separate.
- Report sample count and a distribution, not just one average.
- Use the same exclusions, languages and intent mix for comparisons.