Five choices that change the number
1. Which emails count
The denominator matters more than anything else. Count only the emails that expect an answer from you: a person writing with a question, a request or a decision. Leave out:
- Automated mail: notifications, receipts, out-of-office replies. Many carry an Auto-Submitted header (RFC 3834) or list headers (RFC 2369).
- Newsletters and mailing lists.
- Emails where you were only copied and someone else was asked.
- Thank-yous and confirmations that close a conversation.
Include those and your response time looks worse than it is, or better, depending on whether unanswered mail is counted as missing or as infinitely slow.
2. Business hours or clock hours
An email that arrives at 4 PM Friday and gets a reply at 9:30 AM Monday waited about 65 hours on the clock, and an hour and a half of a 9-to-5 workday. Decide which question you’re asking: “how long did the client wait?” (clock) or “how quickly do we act during the day?” (business hours). Write down the hours and the time zone you used.
3. Median or average
One email that waited a week will drag the average far from the typical case. The median, the middle value, describes the typical email. Report both if you can, and look at the slowest few on their own.
4. First reply or resolution
The first reply stops the clock when you answer. Resolution stops it when the request is done. “Got it, I’ll check with the crew and come back to you” is a fast first reply to a slow resolution. Clients feel both.
5. Per message or per conversation
A follow-up from the same person (“just checking in”) is usually the same request, not a new one. Measuring from their first message is honest; measuring from the follow-up hides the wait.
A worked example
Five client emails in one week, business hours 9 AM to 5 PM, Monday to Friday:
| Arrived | First reply | Clock hours | Business hours |
|---|---|---|---|
| Mon 9:00 AM | Mon 10:00 AM | 1 | 1 |
| Mon 4:00 PM | Tue 10:00 AM | 18 | 2 |
| Tue 11:00 AM | Tue 11:30 AM | 0.5 | 0.5 |
| Wed 10:00 AM | Fri 2:00 PM | 52 | 20 |
| Fri 3:00 PM | Mon 11:00 AM | 68 | 4 |
| Clock hours | Business hours | |
|---|---|---|
| Median | 18 | 2 |
| Average | 27.9 | 5.5 |
The same week reads as “about two working hours” or “more than a day”, depending on the choices. Neither is wrong. What’s wrong is comparing one team’s business-hours median with another’s clock-hours average.
How to measure it by hand
- Export a week or a month of received and sent mail, or take a sample of client threads.
- For each thread, find the first message that asked something of you and your first reply after it.
- Apply the exclusions above.
- Compute the wait in the hours you chose, then the median and the average.
- List the slowest five threads and read them. They usually explain more than the numbers.
Setting a target
Benchmarks from other companies rarely transfer: a law firm, a field-service company and a startup each have different clients and different hours. A target that works is specific: “client emails get a first reply within one business day, and a resolution or a date for one within three”.
How alfred_ uses the same idea
alfred_’s list of replies you owe applies these choices for you. It counts only emails from people that expect an answer, leaves out no-reply senders, newsletters and notifications, and counts working time: a thread joins after one working day without a reply, and the clock pauses from Friday at 6 PM to Monday at 9 AM in your time zone.
