Most contact centers should plan around 4 to 6 closed chats per agent per hour for standard digital channels, which works out to roughly 7 to 9 minutes of concurrent handling per chat. Drop below 3 and you're likely understaffed or drowning in complex cases; push past 8 and quality almost always pays for it. The figure only holds when your chat mix is routine billing and order questions, your agents run two to three concurrent sessions, and "closed" means resolved without a reopen inside 24 hours. Change any of those and the number moves.
That narrow band is where the argument starts, because the chats-per-hour figure is the most over-quoted and least understood metric in digital service. An agent closing 12 chats an hour looks like a star until you learn 4 of those customers reopened the conversation the next morning. A team hitting 5.5 with a 4 percent reopen rate is quietly doing better work.
Here's what the benchmark actually measures, where the credible ranges sit, and how to set a goal your agents can hit without gaming the close button.
What "Closed Chats per Hour" Actually Counts
Closed chats per hour divides the number of chat sessions an agent ends during a defined window by the productive hours in that window, typically excluding breaks, meetings, and offline admin time. ICMI's resource library treats it as a throughput measure, usually filed under contacts handled per hour alongside voice calls. The denominator matters as much as the numerator. If you count login hours but your agent spends 20 minutes a day in team huddles, your rate is understated.
The word "closed" carries more weight than most dashboards admit. A close can mean the agent ended the session, the customer went silent past a timeout, or the system auto-closed after inactivity. Lean Solutions' guidance on chat metrics flags exactly this ambiguity. Those three definitions produce wildly different rates on the same agent, so a goal without a close definition is meaningless.
Digital channels also break the voice-based logic most workforce teams learned. A voice agent handles one conversation at a time; a chat agent may juggle two, three, or more. That concurrency is why chat throughput looks dramatically higher than call throughput and why copying a voice target onto chat fails immediately.
Conversion is simpler. Divide total closed sessions by total productive hours across the group, never by headcount. A ten-agent team that closes 480 chats across 80 productive hours sits at 6.0, regardless of how many people clocked in.
Why the Realistic Band Sits at 4 to 6 per Hour
The 4-to-6 band comes from running the math backward from handle time rather than from any single published standard. At five chats an hour, average handling time per chat is 12 minutes of clock time, but with three concurrent sessions an agent is actively typing for roughly 4 minutes per chat. That aligns with what routine billing, order status, and password-reset conversations demand.
Above six, something has to give. Either handle time drops below a credible floor for your topic mix, concurrency climbs past what quality tolerates, or agents start closing sessions before the customer is done. Below four, you're either taking unusually complex contacts or carrying idle capacity you're paying for.
Efficiency ratios published by benchmark firms shift year to year, and the specific figure you'll see depends on the sample, region, and channel mix. Treat any single published number as directional. The band holds because the underlying arithmetic is stable: conversation length, concurrency, and wrap-up time set the ceiling and floor.
Set the initial target in the middle of the band and adjust after four weeks of your own data. A goal pulled from a generic benchmark without your reopen rate attached is a guess wearing a suit.
The Concurrency Multiplier Most Targets Ignore
Concurrency is the hidden lever. An agent handling three simultaneous chats doesn't triple throughput, because each conversation slows when attention splits. The realistic gain is closer to 1.8 to 2.2 times single-session output depending on topic complexity, a range consistent with workforce management guidance from bodies like the Society of Workforce Planning Professionals. That's why concurrency targets belong in the same policy document as the chats-per-hour goal.
Push concurrency to four or five routine sessions and per-chat response time stretches. Customers notice the lag even when resolution time stays flat. The throughput number rises while the experience quietly degrades.
Complex topics invert the math. A technical support chat involving screenshots and back-and-forth troubleshooting may run 20 minutes and resist concurrency entirely. Mix those into a queue with password resets and the blended rate means little.
Pair every concurrency setting with a response-time ceiling. If your channel promises a first response within 30 seconds, cap concurrency where agents can still meet it. The chats-per-hour goal should follow from that ceiling, not the other way around.
How Channel Mix Moves the Target
Not all digital channels behave alike, and a single blended goal across them hides more than it reveals. Live chat, asynchronous messaging, email, and social DM each carry different rhythm and different definitions of "closed."
| Channel | Typical closed per hour | Handling pattern | Close definition | Main risk to the target |
|---|---|---|---|---|
| Live web chat | 4β6 | Synchronous, 2β3 concurrent | Agent ends session | Reopens within 24 hours |
| Async messaging (in-app) | 6β10 | Burst handling across hours | Thread marked resolved | Customer returns same thread |
| 3β6 | Batch, one at a time | Ticket status set to closed | Reopen rate often untracked | |
| Social DM | 5β8 | Short, high-volume | Conversation archived | Public escalation after close |
| Co-browse or screen share | 2β3 | Fully synchronous | Session terminated | Long tail of follow-up work |
The async row is the one that trips people up. Messaging lets an agent park a conversation, answer three others, and return, so raw closes per hour run higher. Whether that counts as productivity depends on whether customers get answers inside the promised window.
Set one goal per channel and report them separately, then blend only for executive dashboards. A blended figure across live chat and email will hide a failing channel behind a strong one.
Live chat's synchronous nature makes it the strictest measure. An agent can't close a live session while the customer is still typing, so the rate is bounded by real conversation length. Email's batching makes its rate look low on an hourly basis while total daily output stays healthy.
Worked Example: Building a Target From Handle Time
Run the arithmetic before you publish any goal. This keeps every number in your target traceable to something the team can see.
1: Pull the median handle time per chat from your platform for the last 30 days, split by topic. Assume the routine subset averages 9 minutes of clock time per session. 2: Set concurrency at 2.5 routine sessions, then apply a realistic efficiency factor of 0.8 for attention switching, giving an effective 2.0 sessions. 3: Divide 60 minutes by the 9-minute handle time and multiply by the 2.0 effective concurrency: 60 Γ· 9 Γ 2.0 = 13.3 active minutes of session time per hour. 4: Subtract wrap-up, which at 90 seconds per chat across 6.7 closes eats about 10 minutes of the hour. 5: The result lands near 5.5 closed chats per hour, squarely inside the 4-to-6 band.
If the arithmetic lands outside the band, the cause is almost always a data problem rather than an agent problem. Check first whether wrap-up time is being logged. Plenty of platforms close the session timer the moment the customer leaves, so after-chat work never appears anywhere.
Second, split the data by topic. A team whose average hides a complex tier will look slow until the tier is separated out. Third, verify that productive hours exclude meetings and coaching, or the denominator inflates and the rate looks artificially low.
Publish the calculation alongside the goal. Agents who can see where 5.5 comes from argue about the inputs rather than the number, and that argument is more productive.
Quality Guardrails: Reopen Rate, CSAT, and First-Contact Resolution
A throughput goal without guardrails rewards closing fast, and the cheapest way to close fast is to close early. Attach three guardrails to every chats-per-hour target and make them non-negotiable.
Reopen rate is the sharpest. Track the percentage of closed chats where the same customer returns within 24 hours, a window consistent with common quality-monitoring practice. If reopens rise while throughput rises, the goal is being gamed, whether deliberately or not.
CSAT on closed conversations catches the experience cost. A team hitting 6.5 with CSAT slipping two points month over month is trading tomorrow's loyalty for today's dashboard.
First-contact resolution, measured per conversation rather than per contact, closes the loop. These three metrics move together or the throughput number is fiction.
Set a hard ceiling: no individual target above the band's top end unless the agent's reopen rate and CSAT stay at team average. That single rule prevents the fastest closers from setting an unreachable pace for everyone else.
Segmenting Goals by Topic, Tenure, and Tier
A single center-wide target punishes new agents and rewards easy queues. Segment instead, using the same dimensions your routing already understands.
New agents in their first 90 days should carry a goal around 60 to 70 percent of the team rate, reflecting the learning curve documented in contact center onboarding research. Tenured agents on the complex tier shouldn't be measured on the same scale as billing specialists.
Tier the target by topic complexity: routine transactions at the top of the band, troubleshooting in the middle, technical escalations at the bottom or excluded from the hourly measure entirely. Report escalations on resolution time instead.
Tenure segmentation has a nasty edge case. An agent who transfers from a simple queue to a complex one and keeps the old goal will look like a sudden underperformer. Reset the target with the queue change, in writing.
Review segmentation quarterly. Queue mixes drift as products launch and self-service absorbs the easiest contacts, and a stale segmentation quietly penalizes whoever's queue got harder.
Common Ways the Metric Gets Gamed
The metric is easy to game, and agents under pressure find the shortcuts fast. Knowing them is the difference between a target that improves service and one that trains bad habits.
Premature close. Agent ends the session while the customer is still typing. Reopen rate spikes a day later. Silent timeout farming. Agent lets idle timers expire sessions rather than resolving them, then counts the auto-close as throughput. Cherry-picking. Agent skips complex chats in queue and grabs quick ones. Individual rate climbs; queue abandonment rises. Wrap-up skipping. Agent closes the session before logging notes. Throughput looks great; the next agent starts blind. Thread splitting. Agent closes and reopens a conversation to count it twice. Rare, but it happens. Soft-transfer dumping. Agent pushes the customer to another queue and closes. The receiving team absorbs the work.
Each shortcut leaves a fingerprint. Premature closes show up in reopen rate, cherry-picking in queue abandonment, wrap-up skipping in note-completion audits. Build those three reports before you build the throughput dashboard.
Monitoring, Coaching, and Resetting the Goal
A target is a hypothesis, and hypotheses need a review cycle. Start with a 30-day pilot on a subset of agents and compare against a control group on the old measure.
Watch the variance, not just the average. A team averaging 5.5 where half the agents sit at 3 and half at 8 has a coaching problem, not a target problem. Individual spread wider than two chats per hour usually signals a skill gap or a queue-routing issue.
Coach to the inputs. Handle time, concurrency management, and wrap-up discipline are coachable; the hourly rate is an output. A conversation about why a chat took 14 minutes beats a conversation about why the rate was 4.2.
Reset the goal when the inputs change materially: new products, a self-service launch that removes easy contacts, a concurrency policy change, or a platform migration that alters timers. Document the reset reason so the trend line stays readable.
FAQ
What is a good number of closed chats per hour for a contact center?
The working band is 4 to 6 per agent per hour for routine live chat with two to three concurrent sessions. Async messaging and social DMs often run 6 to 10 because agents park and resume threads. Email typically sits at 3 to 6 on an hourly basis even when daily output is strong.
Does higher concurrency always raise chats per hour?
No. Moving from two to three concurrent sessions typically lifts throughput by roughly 1.8 to 2.2 times single-session output, not three times, because attention splits and each conversation slows. Past three sessions, response lag grows and customers notice even when total resolution time holds.
Why do top performers sometimes have worse reopen rates?
Because closing early is the fastest route to a high hourly rate. An agent who ends a session before the customer is finished posts strong throughput and generates a return visit the next day. Track reopen rate inside 24 hours alongside the rate itself so the two numbers are always read together.
Should email and live chat share one chats-per-hour goal?
No. Email batches across the day, live chat runs in real time, and their close definitions differ. Set one target per channel, report them separately, and blend only for leadership summaries, where a single figure is easier to read than to act on.
What happens to the goal when a new product launch adds complex contacts?
The blended rate drops even if every agent performs identically, because the topic mix shifted. Reset the target with the queue change and document the reason. Leaving the old goal in place during a launch guarantees a team-wide miss that says nothing about effort.
Is handle time on the chat timer the same as time to resolution?
Usually not. Most platforms stop the session timer when the customer or agent exits, so after-chat work and any follow-up disappear from the number. Check your platform's timer behavior before you build a target on it, because wrap-up often hides outside the reported handle time.
Do new agents need a lower target during onboarding?
Yes, typically 60 to 70 percent of the team rate for the first 90 days. Ramp time is a documented feature of contact center onboarding, and holding a new hire to the tenured number in month one mostly produces turnover or corner-cutting, sometimes both.
How often should the chats-per-hour goal be reviewed?
Review quarterly and reset whenever the inputs change materially: a concurrency policy shift, a self-service launch, a product change, or a platform migration. Run a 30-day pilot against a control group before rolling a new target out center-wide.