A Lower Average Can Coincide With More Inquiries Waiting Over Two Hours
Average reply time has fallen, yet customers still complain about slow answers. In that situation, examine how many inquiries involve long waits alongside the average. A few exceptionally slow replies may have become faster while more inquiries exceed the time customers need an answer.
The mean is the sum of reply times divided by the number of inquiries. Every long wait contributes to that sum, so a change in a few long waits can change the average. The median is the central value when reply times are ordered, and a few exceptionally slow replies affect it less than they affect the mean. (Measures of Location)
Compare the average across all inquiries and count those exceeding the response time stated to customers. If that count rises despite a lower average, investigate the delays.
Average Reply Time Falls While the Two Hour Reply Rate Declines
Compare two hypothetical sets of ten inquiries.
Assume a help desk records ten inquiries received in each of two intake periods, before and after a change. Wait until every inquiry has received its first answer before compiling the comparison. Measure elapsed time from receipt to the first staff reply addressing the question. Exclude automated acknowledgments and include nights and holidays. This measures the first answer rather than the time to complete resolution.
| Measure | Ten inquiries before the change | Ten inquiries after the change |
|---|
| Reply times | Nine take 1 hour and one takes 11 hours | Six take 0.5 hours and four take 3 hours |
| Sum of reply times | 20 hours | 15 hours |
| Mean reply time | 2 hours | 1.5 hours |
| Median | 1 hour | 0.5 hours |
| Inquiries answered within 2 hours | 9 | 6 |
| Reply rate within 2 hours | 90% | 60% |
| Longest reply time | 11 hours | 3 hours |
The original mean is (9 × 1 + 1 × 11) ÷ 10 = 2 hours. The later mean is (6 × 0.5 + 4 × 3) ÷ 10 = 1.5 hours.
To calculate the reply rate within two hours, divide inquiries answered within that time by all ten inquiries received. Before the change, 9 ÷ 10 = 90%; afterward, 6 ÷ 10 = 60%. The average falls, yet inquiries waiting over two hours rise from one to four. The timely reply rate declines by 30 percentage points.
If customers need an answer within two hours, more inquiries now miss that requirement. Reporting only the 30-minute reduction in the average would hide this problem.
The median and the longest reply time also fall in this example. For ten ordered observations, the median is the average of the fifth and sixth values. Both are one hour before the change and 0.5 hours afterward.
Replacing the mean with the median alone would therefore still miss the increase in inquiries exceeding two hours. Reducing the longest wait from eleven hours to three hours is an improvement, while the proportion answered within two hours declines at the same time. Reporting the average and the number exceeding two hours shows both changes: a shorter maximum wait and more replies missing the two-hour threshold.
Google's engineering book explains how average computer response times can conceal slow responses. For inquiry handling, counting how many answers arrive within the expected time similarly helps reveal the waits customers experience. (Service Level Objectives)
For a help desk stating that answers arrive within two hours, count inquiries exceeding two hours after receipt as late.
Counting Only Answered Inquiries Hides Those Still Waiting
In the reporting table, include answered inquiries and the unanswered count and receipt time of each inquiry. Selecting only those already answered on the reporting date removes ongoing long waits from both the average and the reply rate.
For example, assume that of ten inquiries received, six were answered within two hours, two after three hours, and two remain unanswered four hours after receipt. The reply rate within two hours is 6 ÷ 10 = 60%. Using only the eight answered inquiries omits the two still waiting and reports 75% instead of the actual 60%.
The final reply time for the two unanswered inquiries cannot be entered as four hours. All that is known is that each has already waited four hours. The final mean across all ten is still unknown. However, once every inquiry has passed the two-hour mark, whether each received an answer within two hours can be determined.
Group inquiries by the period in which they arrived and calculate the timely reply rate after each has had the full threshold time to receive an answer. Enter unanswered inquiries that have not yet passed the threshold in a pending-assessment column. Calculate each current wait by subtracting receipt time from reporting time, and enter those waiting times in the unanswered-inquiry list.
Report Average Time Alongside Replies Missing the Stated Response Time
In the example, waiting time per inquiry fell by 30 minutes on average. If the objective is to reduce waits exceeding two hours, however, the example moves against that objective. Compare the average, timely reply rate, and unanswered count using the same intake periods and time definitions.
Adding automated acknowledgments to a measure previously based on staff answers, or switching from elapsed hours to business hours, changes the figures even without a change in operations. If difficult questions or overnight arrivals have increased, examine those categories to find where delays are concentrated.
On your website, state opening hours, closed days, and whether a response estimate refers to the first answer or completed resolution. Customers can then distinguish business-day counting from elapsed time and know whether to expect an initial answer or a resolved problem. Avoid presenting a historical average as a deadline that applies to every inquiry.
SpreadSite is a no-code website builder that lets you create, publish, and update a website from Google Sheets using ready-made templates.
To create a website explaining your inquiry hours and expected answer times, give SpreadSite a try!
Build a website with SpreadSite →
