Average handle time (AHT) is the average amount of time an agent spends actively working on one customer conversation, including the notes and lookups after it. The formula is total handle time divided by the number of conversations handled. For calls that means talk, hold and after-call work. For chat and email it means the minutes someone actually spent reading, investigating and replying. A small team should use it to plan staffing and to spot broken processes, and should stop pushing it down the moment resolution or satisfaction starts to slip.
What average handle time measures
AHT counts agent effort only. The clock runs while an agent is working on the customer's problem and stops while the conversation sits in a queue or the customer goes quiet. That makes it different from the two clocks small teams usually watch:
| Metric | What it measures | Clock runs while | Mostly useful for |
|---|---|---|---|
| Average handle time | Agent effort per conversation | An agent is actively working on it | Staffing, spotting slow processes |
| First response time | How long the customer waits for the first reply | The conversation is unanswered | Customer experience, service levels |
| Resolution time | How long the issue takes to close, end to end | The conversation is open, including waiting on the customer | Customer experience, backlog health |
An email ticket can have a handle time of 9 minutes and a resolution time of three days, because most of those three days were spent waiting for the customer to send a screenshot. Mixing the two up is the most common AHT mistake on email teams.
The average handle time formula, with examples
The general formula is the same everywhere:
AHT = total handle time ÷ number of conversations handled
What goes into "total handle time" changes by channel.
Phone
For calls, the standard version is (talk time + hold time + after-call work) ÷ calls handled. If an agent takes 40 calls in a day with 160 minutes of talk, 20 minutes of hold and 60 minutes of wrap-up, AHT is 240 ÷ 40 = 6 minutes.
Email tickets often take several touches, so add every touch on a ticket before you divide. Here is one hypothetical agent's morning:
- Ticket A, refund request: 5 minutes to reply, then 3 minutes to process the refund after the customer confirms. Handle time 8 minutes.
- Ticket B, "how do I export my data?": 4 minutes, one reply with a help article. Handle time 4 minutes.
- Ticket C, bug report: 15 minutes reproducing it, 4 minutes writing it up for the developer, 3 minutes updating the customer the next day. Handle time 22 minutes.
- Ticket D, change of billing email: 2 minutes. Handle time 2 minutes.
Total handle time is 8 + 4 + 22 + 2 = 36 minutes over 4 tickets, so AHT is 9 minutes. Note that the median ticket took 6 minutes. One bug report pulled the average up by half, which is why the spread matters as much as the average.
Live chat
Chat is where AHT gets misread, because agents handle several chats at once. If you average the length of each chat window, you count the overlapping minutes twice.
Suppose an agent handles 30 chats in a three-hour shift and averages two open at a time. The chat windows add up to 360 minutes, an average of 12 minutes per chat. The agent only worked 180 minutes, though, so the effort per chat is 180 ÷ 30 = 6 minutes. Both numbers are real. The 12 minutes describes the customer's experience; the 6 minutes is what you need for staffing. Decide which one your reports call AHT and write the definition down.
Where the numbers come from
Phone systems record handle time automatically. Many help desks do not record active time on email and chat, or record "time the ticket was open in a tab", which is not the same thing. If yours does not, time 50 conversations by hand over a week, split by ticket type. That sample is enough to plan with, and you will learn more from doing it than from any dashboard.
What a normal average handle time looks like
You will see "about six minutes" quoted as a good AHT. Zendesk's AHT guide gives that figure, plus industry estimates such as 3 to 4 minutes for retail and 8 to 10 for technical support, without naming a source. Those are call center numbers, measured with call center definitions.
For a small team on chat and email, there is no public benchmark worth targeting. Teams disagree on whether wrap-up counts, how to treat concurrent chats and whether a multi-touch email is one handle or several, so two teams with identical work can report numbers that differ by half. And your mix of questions decides the result far more than how fast anyone types.
A better baseline is your own, split by ticket type. A hypothetical SaaS team might find:
| Ticket type | Median handle time | 90th percentile |
|---|---|---|
| Password or login help | 3 minutes | 6 minutes |
| Billing question | 6 minutes | 15 minutes |
| How-to question | 7 minutes | 18 minutes |
| Bug report | 18 minutes | 45 minutes |
Those numbers tell you something useful. If billing questions suddenly take 12 minutes at the median, something changed in billing. A single blended AHT would hide that.
Why lowering average handle time can backfire
AHT is easy to improve in ways that make support worse. Here are the patterns to watch for:
- Rushed answers come back. A 4-minute reply that misses the real question produces a second ticket, and the customer now has to explain twice.
- Transfers stop the clock. On teams measured on AHT, handing a hard conversation to someone else looks like efficiency while the customer waits longer.
- Notes get skipped. Wrap-up time is part of AHT, so it is the first thing people cut. The next agent then spends ten minutes rebuilding the context.
- People pick the easy tickets. If handle time shows up in reviews, the bug reports sit at the bottom of the queue.
- Good changes can push it up. Self-service removes your quickest tickets, so the average of what is left rises.
Here is that last effect with numbers. Say a team handles 200 password tickets at 3 minutes and 100 bug reports at 18 minutes in a month. AHT is (600 + 1,800) ÷ 300 = 8 minutes. The team then adds a self-service reset link and password tickets fall to 80. Next month AHT is (240 + 1,800) ÷ 180 = 11.3 minutes. Nobody got slower, the team saved six hours, and the headline AHT went up 42%.
The fairer number is time per resolved issue: handle time multiplied by how many contacts an issue takes. An agent averaging 6 minutes with 1.6 contacts per issue spends 9.6 minutes per problem solved. A colleague at 8 minutes with 1.1 contacts spends 8.8, and their customers only had to write once.
When to watch AHT and when to stop chasing it
AHT is useful for three jobs on a small team:
- Staffing. Monthly conversations × AHT gives the hours of work. 1,200 conversations at 7 minutes is 140 hours, before breaks, meetings and the time people spend between conversations. The free agent capacity planner turns that into a headcount.
- Finding broken processes. When one ticket type takes three times longer than it used to, the cause is usually a missing tool, a confusing policy or a product bug, not a slow person.
- Coaching new hires. Comparing a new agent's handle time per ticket type with the team median shows where they need help, as long as quality is reviewed alongside it.
Stop chasing it when any of these happen: AHT becomes a target in performance reviews, first contact resolution or CSAT starts to fall, reopen rates climb, or your ticket mix is shifting because of self-service. In each case, a lower number is no longer evidence of better support.
Keep it on the dashboard as a diagnostic next to first contact resolution, reopen rate and CSAT. Our support analytics page covers how those fit together, and the support KPI calculator works them out from numbers you already have.
Ways to cut handle time without hurting quality
The safe gains remove friction for the agent instead of pressure on them:
- Saved replies for your top ten questions, written well once and edited per customer. The canned responses examples give you a starting set.
- Customer context beside the conversation: past tickets, plan, order status. Most lookup time is switching between tabs. A shared inbox that shows history next to the thread removes most of it.
- Internal notes and mentions in the conversation itself, so a question to a colleague does not turn into a Slack thread nobody can find later.
- Routing that gets a conversation to the right person first. Every transfer adds a full read-through. Simple rules, like billing to the person who has billing access, are covered in our guide to customer service automation.
- Better help articles, linked in replies. A good article shortens the reply and stops the follow-up question.
- Fixing the cause. If one ticket type takes 18 minutes because reproducing a bug is slow, a debug log or an admin view will save more time than any amount of coaching.
- Letting AI take the routine questions. An AI agent on website chat that answers from your help content and hands over when unsure removes the shortest tickets, which raises the blended AHT for the reasons above. Judge it by hours saved. Sellio's AI agent works this way on website chat, with a cap on spend.
FAQ
What is a good average handle time?
Call center guides often quote about six minutes, with industry estimates running from 3 to 10. For a small team on chat and email, a good AHT is one that is stable for each ticket type while first contact resolution and CSAT hold steady. Compare against your own baseline, not someone else's average.
Does average handle time include customer wait time?
No. Queue time and time spent waiting for the customer are excluded. Hold time on a call is the exception, because the agent is still occupied with that customer. Wait time shows up in first response time and resolution time instead.
How do you calculate AHT for email?
Add up the active minutes for every touch on each ticket (reading, investigating, writing, wrap-up), then divide the total by the number of tickets. Count a ticket once, however many replies it took.
Is average handle time the same as resolution time?
No. Handle time is agent effort. Resolution time is the elapsed time from opening to closing, including all the waiting in between. A ticket can take 10 minutes of effort and three days to resolve.
Should small support teams track AHT at all?
Yes, loosely. Measure it per ticket type a few times a year, use it for staffing and to spot slow processes, and keep it out of individual targets. First contact resolution and CSAT tell you more about whether customers were helped.
