Measure average time from question to answer

Response-time measurement should use consistent start and end events so queue delay is not confused with voice generation or playback delay.

A practical procedure

1. Choose the start timestamp, such as question arrival, and the end timestamp, such as first audible answer word.

2. Calculate elapsed time for each answered question using the same clock.

3. Report the average with the range and list unanswered questions separately.

Worked example: a fictional session

A fictional sample has delays of 2, 4 and 6 seconds. Its average is 4 seconds, while the six-second case remains important to inspect.

Check the limits

Excluding unanswered questions from timing is reasonable only when their count is reported too.