What is average handle time?
Average handle time (AHT) is the mean amount of active time a support agent spends on a single ticket, from first read to close, including any follow-up and after-contact work. It is a capacity metric, not a satisfaction metric: a lower average handle time means each agent clears more tickets per hour, which feeds directly into staffing levels, queue depth, and cost per ticket.
For ecommerce specifically, AHT behaves differently than it does in a call center. Most ecommerce volume is asynchronous — email, chat, WhatsApp, Instagram DM — so a single ticket can stay open across hours while the agent touches it for only a few minutes. That is why ecommerce AHT measures active handling time, not the elapsed time the ticket sat in the queue. Confusing the two is the most common reporting mistake we see.
Treat AHT as a per-category number, not a single blended average. An 8-minute blended AHT can hide a 22-minute average on damaged-item disputes that is quietly burning a third of your team's week. The blended figure tells you what staffing costs; the breakdown tells you where to fix things.
AHT also sits at the center of every staffing model you build. If your team handles 1,200 tickets a week and your blended AHT is 10 minutes, that is 200 agent-hours of pure handling time before you account for breaks, training, or queue gaps. Shave that AHT to 7 minutes and you reclaim 60 hours a week — roughly a full-time agent — without losing a single resolution. That is why AHT, unglamorous as it is, drives more headcount decisions than any other support metric.
Average handle time = total active handling time across all tickets ÷ number of tickets resolved. In ecommerce it counts the minutes an agent is actively reading, researching, replying, and wrapping up — not the hours a ticket waits unattended in the queue.
How to calculate AHT
The classic call-center formula adds three components and divides by ticket volume: talk/handling time, hold time, and after-contact work. For ecommerce email and chat there is rarely a literal hold, so the formula simplifies to active handling time plus wrap-up time, divided by the number of tickets.
The part teams get wrong is wrap-up. After-contact work — tagging the ticket, updating the order note, logging a refund, writing an internal handoff — is real handling time and belongs in AHT. Leave it out and your number looks better than reality, which leads to understaffing. Most help desks capture handling time automatically, but only if agents actually close tickets instead of letting them linger open.
One more reporting trap: averages hide outliers. A handful of 90-minute fraud investigations will drag a blended mean upward and make a healthy team look slow. Report the median alongside the mean, and segment by ticket type, so a few painful edge cases don't distort the picture you use to plan headcount.
There is also a difference between AHT and resolution time that trips up new managers. Resolution time is the full lifespan of a ticket — pickup to close — and on an email ticket that waits a day for the customer to reply, it can read as 26 hours. AHT counts only the minutes your agent was actually working. A ticket can have a long resolution time and a short AHT, and that is normal and healthy for asynchronous channels. Conflate the two and you will conclude your team is slow when it is simply waiting on customers.
| Component | What it includes | Counts toward AHT? |
|---|---|---|
| Active handling time | Reading, researching, drafting, replying | Yes — the core of AHT |
| After-contact work | Tagging, order notes, refund logging, handoffs | Yes — often the most underreported part |
| Hold / pause time | Waiting on a system or a colleague mid-ticket | Yes, when the agent is committed to the ticket |
| Queue / wait time | Time the ticket sits unassigned before pickup | No — that is first response time, not AHT |
| Customer reply lag | Hours a thread waits on the customer to respond | No — exclude it from active handling time |
AHT benchmarks by ticket type
The single most useful AHT view is by ticket type, because handle time varies more across categories than it does across well-run teams. Order-status questions are quick and formulaic; damaged-goods disputes are slow and investigative. Your blended AHT is mostly a weighted average of your ticket mix, so two competent teams can post very different numbers purely because one sells simple consumables and the other sells fragile, high-consideration goods.
Industry benchmarks for ecommerce support consistently land AHT in the 5–15 minute range overall, with strong performers clustering at 5–8 minutes. The table below breaks that down by the categories that dominate most stores. Treat the ranges as orientation, not targets — your own per-category baseline matters more than any external figure.
Notice where the leverage sits. WISMO (where-is-my-order) tickets have low handle time but enormous volume, so automating them frees hours through sheer count. Complex disputes have low volume but punishing handle time, so process fixes there save disproportionate minutes per ticket. You attack those two categories with completely different tools.
| Ticket type | Typical AHT | Strong-performer AHT | Share of volume |
|---|---|---|---|
| Order status / WISMO | 3–6 min | 2–4 min | 30–40% |
| Return eligibility questions | 5–9 min | 4–6 min | 10–15% |
| Shipping issues / delays | 7–12 min | 5–8 min | 10–15% |
| Pre-sale product questions | 5–10 min | 4–7 min | 10–20% |
| Billing / refund requests | 8–15 min | 6–10 min | 5–10% |
| Account / login issues | 6–11 min | 4–8 min | 5–8% |
| Damaged / wrong item disputes | 15–30 min | 12–20 min | 3–8% |
| Overall blended AHT | 6–15 min | 5–8 min | 100% |
Across ecommerce email and chat, AHT typically runs 5–15 minutes per ticket. Simple, data-grounded contacts (order status, basic product questions) average 3–6 minutes; investigative tickets (disputes, multi-item returns) average 12–25+ minutes. Strong teams hold a blended 5–8 minutes.
AHT benchmarks by channel
Handle time also shifts by channel, and the differences are easy to misread. Live chat looks fast per message but eats agent attention because it is synchronous and often concurrent. Email looks slow per ticket but lets one agent batch through a queue efficiently. Social and messaging channels sit in between, with informal back-and-forth that can stretch a thread out.
Phone and voice carry the highest handle time of any channel because they are one-to-one and real-time, with no batching and a mandatory wrap-up. Most ecommerce brands keep voice for high-value or complex cases for exactly this reason. The benchmarks below are directional ranges for active handling time per resolved contact.
The practical takeaway: don't compare a chat-heavy team's AHT against an email-heavy team's and conclude one is more efficient. Normalize by channel mix first, then by ticket type, before you judge performance or set goals.
| Channel | Typical AHT per contact | Why it lands there |
|---|---|---|
| 5–12 min | Asynchronous; agents batch, but threads can run long | |
| Live chat | 8–15 min active | Synchronous and often concurrent across 2–3 chats |
| WhatsApp / Instagram / Messenger | 6–12 min | Informal, multi-turn; context spread across messages |
| SMS | 4–8 min | Short, transactional; limited scope per thread |
| Phone / voice | 10–20 min | Real-time, one-to-one, plus mandatory after-call work |
What drives AHT up
High AHT almost always traces back to one of four root causes: information the agent has to go hunt for, policies too vague to act on without approval, ticket types that are inherently complex, or workflow friction from juggling tools. None of these are about agents working slowly — they are about the system making fast work impossible.
Diagnose before you fix. Sit with three or four high-AHT tickets and time where the minutes actually go. In our experience the surprise is consistent: the reply itself is fast, but the lookup and the decision-making around it are slow.
- Tool-switching: every system an agent leaves the conversation to check adds 1–3 minutes.
- Missing knowledge: agents pause to find a policy that should be one click away.
- Escalation loops: tickets bounced between tiers carry handling time at every hop.
- Manual data entry: re-keying order numbers and addresses instead of pulling them automatically.
Information lookup time
Agents who tab between the help desk, Shopify admin, the carrier site, and a returns portal to assemble one answer spend most of their handle time gathering context rather than resolving anything. A unified inbox that surfaces order, fulfillment, and customer history in the same view as the conversation removes that tax outright.
Policy ambiguity and approval chains
When an agent cannot decide without checking a manager, AHT balloons — and balloons further when the manager is busy. Documented policies plus real authority (issue refunds up to a set cap, approve returns inside the window) collapse a three-touch ticket into one.
Response drafting from scratch
Writing every reply fresh is slow and inconsistent. A maintained template library for your top 15–20 ticket types is one of the highest-ROI moves available, and it compounds: agents reuse, refine, and standardize the language over time.
Ticket complexity mix
Blended AHT is sensitive to your mix. If disputes are 6% of volume but 30% of agent-time, that is where leverage lives — either resolve the upstream cause (damaged goods means a packaging or carrier problem) or systematize the investigation so it isn't reinvented each time.
Why lower AHT is not always better
Cutting AHT is the right goal only when resolution quality holds. Push handle time too low and agents start rushing — sending terse, incomplete answers that generate a second contact within a day. You shaved two minutes off one ticket and created a whole new one. That is why AHT should never be optimized in isolation.
The metric that keeps AHT honest is first-contact resolution. If AHT drops while FCR slips and repeat-contact rate climbs, you have not gained efficiency — you have shifted work downstream and hidden it. Read AHT next to FCR, repeat-contact rate, and CSAT, and treat any AHT win that drags those down as a false positive.
There is a healthy way to lower AHT and an unhealthy way, and they are easy to tell apart. The healthy way removes work — better tooling, clearer policies, AI handling the lookups — so the agent reaches the same quality answer in less time. The unhealthy way removes care — capping reply length, penalizing long tickets, rewarding speed alone — so the answer gets thinner. The first compounds into a better operation; the second quietly inflates your contact volume two weeks later. When you set an AHT goal, be explicit about which kind of improvement you are asking for.
An AHT target enforced on its own quietly trains agents to deflect rather than resolve. Pair every AHT goal with a first-contact-resolution floor and a CSAT floor, so speed never comes at the cost of getting it right the first time.
How AI reduces AHT
AI moves average handle time through two distinct mechanisms: deflection, which removes tickets from the human queue entirely, and augmentation, which makes agents faster on the tickets that remain. They work in opposite directions on the per-ticket number, which is why AHT can look misleading after you deploy AI if you don't account for both.
Deflection has an indirect — and slightly counterintuitive — effect. When an AI agent resolves the easy 50% of contacts (order status, return policy, sizing), the humans are left with the harder half. Their per-ticket AHT actually rises, because every remaining ticket is genuinely complex. But total agent-hours fall sharply, and that is the number that drives cost.
Augmentation works in the other direction and shows up directly. AI drafts a first reply for the agent to review, surfaces the relevant order data in the same pane, suggests the policy that applies, and links similar resolved tickets. Each of those removes a lookup or a drafting step, so the agent spends their minutes on judgment and tone rather than research.
This is also why you should not judge an AI deployment by the human AHT number in isolation. If you read only that figure after launch, it may look like the AI made your team slower — the average ticket is harder now. The honest measure is total handling hours across humans and AI combined, divided by total resolved contacts. That blended cost-per-resolution is what falls, often substantially, even when the per-agent AHT on remaining tickets ticks up. Set up your reporting to show both numbers side by side before you go live, so nobody panics at the wrong one.
| Mechanism | Effect on AHT | How it works |
|---|---|---|
| AI deflection (ticket fully resolved) | Removes the ticket from human time entirely | The agent resolves it end to end — no human minutes spent |
| AI draft reply (human reviews and sends) | 30–50% per-ticket reduction | Editing a strong draft is far faster than writing from scratch |
| Context surfacing (order data in view) | 15–25% per-ticket reduction | Kills the system-switching tax on information lookup |
| Policy and next-step suggestion | 10–20% per-ticket reduction | Agent stops hunting for the applicable rule |
How Bookbag cuts handle time
Bookbag is an AI support agent built for ecommerce, and it attacks AHT from both directions at once. It resolves high-volume, low-complexity tickets autonomously — order tracking, returns, exchanges, refunds within your rules, product questions — so those never reach a person at all. Industry-typical deflection for well-tuned ecommerce AI runs up to around 70% of contacts, and every deflected ticket is handle time you never spend.
For the tickets that should reach a human, Bookbag hands them off with full context already assembled: a conversation summary, the live order and fulfillment data, the relevant policy, and a suggested action. The agent opens the ticket and is ready to decide, not to dig. That is the augmentation half of the equation, and it is where the per-ticket AHT savings on remaining work come from.
Because Bookbag connects natively to Shopify, WooCommerce, and BigCommerce and works across chat, email, WhatsApp, Instagram, and Messenger, the context follows the customer across channels — so an agent isn't reassembling the story every time the conversation hops. Pricing is flat monthly plans with message-credit allowances, not per-resolution, so deflecting more tickets never inflates your bill.
Deflect the easy half autonomously and hand off the hard half pre-researched, and your team stops spending its day on lookups. The minutes that remain go to the few tickets that genuinely need human judgment.
How to lower AHT without AI
You can cut AHT meaningfully before you automate anything, and you should — process fixes make any future AI deployment work better too. These tactics are ordered by return on effort, so start at the top and work down.
The theme across all of them is the same: stop making agents leave the conversation to do their job. Every lookup, every escalation, every blank-page reply is a place where minutes leak.
- 1Build a response-template library for your top 20 ticket types. Every recurring question gets a polished, editable template. This alone typically trims AHT by 20–30% and improves consistency.
- 2Surface order data inside the help desk. Through a native integration or a sidebar, the agent should see the order, fulfillment status, and customer history the moment they open the ticket — never tab over to admin.
- 3Document policies so agents can answer without escalating. Each policy question that needs a manager adds 5–15 minutes; write the rules down and they evaporate.
- 4Give agents real action authority. An agent who can refund up to a set cap and start a return without sign-off resolves in one touch instead of three.
- 5Review AHT by ticket type every month and aim your improvement effort at the highest-AHT, highest-volume category — that is where total time is being burned.
- 6Cut context-switching with a single-pane help desk that keeps conversation, order data, and knowledge base together, so no ticket requires a tour of four tabs.
How to track AHT correctly
Tracking AHT well is mostly about segmenting it and pairing it with quality metrics. A single blended number reported monthly tells you almost nothing actionable; the same number broken out by ticket type and channel, read next to FCR and CSAT, tells you exactly where to invest.
Set baselines per category before you set targets. A 12-minute average on disputes is fine; a 12-minute average on order-status questions means something is broken in your lookup workflow. Without the segment-level baseline you can't tell which is which, and you'll chase the wrong fixes.
Finally, watch AHT as a trend after any process or tooling change, not as a one-time snapshot. A template rollout or an AI deployment shifts the ticket mix, so expect the blended number to move for reasons that have nothing to do with agent speed. Interpret the change against the mix, not in a vacuum.
- Segment AHT by ticket type and channel — never report only the blended figure.
- Pair AHT with FCR, repeat-contact rate, and CSAT so speed never hides quality loss.
- Report median alongside mean to keep outlier investigations from distorting the picture.
- Re-baseline after AI or template changes, since the ticket mix shifts underneath the average.
- Tie the highest-AHT, highest-volume category to a named owner each quarter.
Key takeaways
- Ecommerce AHT benchmarks: 5–15 minutes typical overall, 5–8 minutes for strong performers.
- AHT varies more by ticket type than by team — WISMO runs 3–6 min, disputes 15–30 min.
- Always count after-contact work in AHT, and never confuse active handling time with queue wait time.
- Lower AHT is only a win if FCR and CSAT hold — pair every AHT goal with a quality floor.
- AI cuts AHT two ways: deflecting easy tickets entirely and pre-researching the hard ones for agents.
- A response-template library plus order data in the help desk is the highest-ROI AHT fix without AI.