A first response time target for customer support works best as one promise per channel, written as a share of conversations answered within a set time, with the time taken from what your team already manages on an ordinary week. If you want somewhere to start testing, try this: a person replies to website chat within 3 minutes while chat is switched on, to email within 4 business hours, and to slower channels such as contact forms or a Discord server within one business day.
Those numbers are guesses until you hold them up against your own volume and rota, which takes about an hour with the method further down. We would also write the target in one specific form:
"90 percent of chats get a reply from a person within 3 minutes during staffed hours." The percentage admits that some conversations will miss, and "during staffed hours" stops a Saturday email from wrecking Monday's report, and that one change explains a lot of reports that look worse than the service really is.
Decide what you are timing before you pick a number
First response time runs from a customer's first message to your first reply. It sounds settled until two people on the same team calculate it and get answers an hour apart. Write down four choices first:
- When the clock starts. Use the moment the message arrives, not the moment someone assigns it. Time spent unassigned is exactly the delay you want to see.
- What stops it. Count the first reply from a person. Tools disagree on this: Zendesk's SLA documentation says a public autoreply fulfils its first reply time metric, so a trigger that says "thanks, we got your message" can make a slow queue look instant. If your tool works that way, keep a separate first human reply figure.
- Which hours count. Business hours for email and tickets, the hours chat is switched on for chat.
- How you summarise it. The share within target, the median, and the time under which 90 percent of first replies land.
Leave the average out. Picture ten chats in a morning: nine answered within a minute, and one that waited 51 minutes because it arrived during a team call. The average comes to 6 minutes, which matches nobody's experience. The median is 1 minute and the share within 5 minutes is 90 percent, and together those describe the morning properly: quick, with one miss worth reading.
Calendar time causes the other common distortion. An email sent at 6:30 pm on a Friday and answered at 9:30 am on Monday has waited 63 hours on the clock, or 30 minutes of a 9 to 6 working week. Report it in calendar hours and every Monday looks like a failure, so people stop opening the report.
Give each channel its own target
Someone typing into your chat widget is on your site right now, often halfway through a purchase or stuck on a setup screen. Someone who emailed has gone back to their day and will look again later. Hold both to one number and it ends up too slow for chat, or it chains a person to the email queue for no good reason.
The useful question is what the customer is doing while they wait. Start from that and treat the right-hand column as something to test:
| Channel | What the customer is doing | Starting target to test |
|---|---|---|
| Website live chat | Sitting on the page, deciding whether to stay | 2 to 5 minutes while chat is on |
| Email and contact forms | Getting on with their day | 4 business hours, or the same business day |
| Telegram and similar messaging apps | Expecting something between chat and email | 15 to 60 minutes during staffed hours |
| Discord or a shared Slack channel | Asking in public, often with others reading along | 1 to 2 hours during staffed hours |
| Urgent, on any channel | Losing money or access: a double charge, a site that is down, a locked account | 30 minutes during staffed hours |
Four or five targets is about the limit for a team under ten people. Every extra tier is another thing to remember and another argument about which tier a message belongs in, so define urgent in plain words with examples and keep it to one tier.
Benchmark roundups usually put a strong live chat reply under a minute and good email replies within about four hours. That describes large teams with people whose only job is the queue. It is useful context, and a poor target for three people who also pack the orders.
Working out a first response time target for customer support from your own numbers
Most targets fail the same way: someone suggests a round number in a meeting, the team misses it every Monday, and the report quietly stops being read. An hour with real data avoids that. For each channel, collect:
- Conversations in a typical week.
- Conversations in your busiest hour, averaged over four weeks if you can.
- The hours someone is genuinely free to reply, which are often shorter than your opening hours.
- Minutes of attention one conversation takes, from first look to closing it. Ask the team, then time a few.
- Your current median first reply.
- The time under which 90 percent of first replies land today.
If you can't get the last two, measure for two weeks before you promise anything. A spreadsheet with "received at" and "first replied at" columns is enough. Once you have counts, the free first response time SLA estimator turns replies within target and total replies into an attainment rate, and tells you how many more on-time replies it would take to reach 95 percent.
A rough staffing check
For your busiest hour, multiply conversations by minutes of attention, divide by how many chats one person can handle at once, then divide by 60. The result is roughly how many people that hour needs. It ignores breaks, bunching and the one customer who takes 40 minutes, so read it as a floor. Arrivals come in clumps, and once people are close to fully busy, waits stop growing gently and start growing fast. If your chat target is measured in minutes, plan real slack into the peak. The support agent capacity planner does the same sums if you would rather not.
A hypothetical shop with three support people
The shop is staffed 9 to 6 on weekdays and gets about 300 chats and 200 emails a week. Chat peaks between noon and 2 pm at 18 an hour. A chat takes about 8 minutes of attention and each person can juggle two, so lunchtime needs 18 × 8 ÷ 2 ÷ 60 = 1.2 people on chat. Allowing for clumps, call it two.
That leaves one person for email. Around 40 emails a day at roughly 4 minutes each is a bit under three hours of work, which is comfortable for one person as long as they are not also covering chat over lunch.
Say their median chat reply is already 2 minutes, but the slowest tenth wait 14 minutes or longer, nearly all of them between 12 and 2. Reasonable targets would be 90 percent of chats answered by a person within 3 minutes while chat is on, 90 percent of emails within 4 business hours, and urgent order problems within 30 minutes on any channel. The gap closes with a second person on chat from 12 to 2, and nobody has to be hired.
Two founders doing support between everything else
Picture a small SaaS company where the two founders answer about 20 chats and 25 emails a day. Peak chat is 4 an hour, and the staffing check says a quarter of a person covers it. What the check can't see is that neither founder sits in front of the inbox for a whole day.
So they promise less. Chat shows as online from 10 to 2, with a 5 minute target inside that window. Outside it, the widget asks for an email address and says to expect a reply the next business morning. Email gets "by 11 am the next business day" for anything sent before 5 pm. That is slower than the shop's promise, and it is one they can still keep in a week when both of them are buried in a release.
One real team's numbers
The support team at Fyle wrote up how three people took their average first response from about an hour down to 16 minutes. Tickets had been handled first come, first served, so at any given moment nobody in particular owned the queue. Their first change was a shared support calendar, which later became proper shifts, and in the first week of that pilot the average dropped to 21 minutes with the same three people working at the same pace.
Choosing the number
Put the staffing check next to what you already do. Pick a time you currently hit around 70 to 80 percent of the time and set the target at 90 percent within it, so getting there takes a real change, like a lunchtime rota or a named owner, rather than heroics. A target the team already misses half the time only teaches them that targets are decoration. After two months of hitting it, tighten the time or the percentage, one at a time.
Teams spread across time zones should set targets per coverage window. A team split between Lisbon and Toronto might promise 3 minutes on chat from 8 am to 8 pm Toronto time and make no chat promise outside that, instead of stretching one number over hours when a single person is awake.
When the arithmetic says you are short
Sometimes the staffing check comes back with more people than you have. You can switch chat on for fewer hours, loosen the target, or add people. Try the first before the third. A smaller promise you keep does less damage than a bigger one that breaks every busy afternoon.
If you do add people, look at how your software is billed first. Per-seat pricing turns a lunchtime hire into a bigger tool bill as well as a salary: Zendesk's published pricing lists Suite Team at $55 per agent a month, so five seats cost $275 a month (prices checked 2026-09-06). Sellio is priced per workspace, so its Mini plan is $20 a month for up to 5 seats, and a fifth person adds nothing to the bill until you pass that limit. Every plan is on the pricing page, and the seat vs workspace pricing calculator runs the comparison for your own headcount.
Keeping the target once it is set
Most missed targets trace back to messages nobody thought were theirs, or messages nobody saw at all. A handful of habits covers most of it:
- Name one person as chat owner for each block of the day, with a backup. When everyone keeps half an eye on chat, nobody is watching it.
- Assign email round robin, or route billing and technical questions by tag if you have specialists.
- Tag messages that mention things like "charged twice", "can't log in" or "site is down" as urgent, and send them to whoever is on call.
- Warn at about 75 percent of the target: around 2 minutes into a 3 minute chat target, 3 hours into a 4 hour email target. Send the warning where the team already works. If they live in Slack, a ping there gets seen.
- Start each morning by clearing the overnight queue, oldest first.
Getting every channel into one shared inbox helps more than any single rule, because a Telegram message can't sit forgotten in an app nobody opened today. The usual customer service automation patterns for tagging, assignment and reminders are worth setting up once and then leaving alone.
Auto-replies need a rule of their own. An acknowledgement that says "We reply to email within 4 business hours. If it's urgent, put URGENT in the subject" is useful, because it turns an open-ended wait into a known one. One that exists only to stop a clock misleads your own report. Send one per conversation and keep it out of the first human reply figure.
Where AI on website chat changes the numbers
An AI agent can answer at 3 am within seconds, and it can be wrong within seconds too. If AI replies count as first responses, your chat figure will look superb and tell you very little, so track two numbers: time to any first reply, which is what the customer experiences, and time to the first human reply on conversations that reached a person, which shows whether handoffs disappear into a queue.
Sellio's AI agent answers on website chat from your own site and documents, hands over to a person when it is unsure, and can be capped so spend can't run away (usage past the included credits costs extra). It does not reply on email, Telegram or any other channel, so the targets for those rest entirely on your team. Give handoffs their own rule: during staffed hours the human chat target counts from the moment of handoff, and outside them the customer is told when someone will be back.
AI suits documented, repeated questions such as shipping times, plan limits and password resets, and it can cover the hours when the widget would otherwise say you are away. Refunds, exceptions, upset customers and anything touching account security belong with a person. For the first month, read a sample of AI conversations every week. A confident wrong answer usually points at a gap in your docs, and fixing the doc tends to improve speed and accuracy together.
Reviewing response time, resolution and CSAT each week
First response time on its own rewards quick, empty replies, so read it beside resolution time and CSAT. Sellio's support analytics show all three over the same conversations as the inbox, and at low volume a spreadsheet does the job.
Fifteen minutes a week is plenty. For each channel, look at the share that met target, the median and the slowest tenth, median resolution time, CSAT with the number of responses behind it, and what is still open. Then open the single worst miss of the week and read the whole conversation, which usually explains more than the chart does.
| What you see | What it usually means | What to try |
|---|---|---|
| Fast first reply, low CSAT | Replies are quick but don't help | Read the low-rated conversations, then fix docs and saved replies |
| Fast first reply, slow resolution | The first reply is a holding message, or the fix sits with another team | Look at handoffs and time spent waiting on someone else |
| First reply slipping as volume rises | Not enough people at certain hours | Check the peak hours, move shifts or shorten chat hours |
| One channel slow, the rest fine | Nobody owns that channel | Name an owner and add a warning |
| Email slow, CSAT fine | Customers accept that pace there | Consider loosening it and spending the effort elsewhere |
| Every target met, team worn out | Targets are being hit with overtime | Cut chat hours or add help before someone burns out |
Two cautions. With 20 CSAT responses a month, one unhappy customer moves the score by 5 points, so read the comments before you react to the percentage. And don't rank individuals on first response time. People ranked on speed learn to send fast, thin replies, which is the failure all of this is meant to prevent.
Revisit the targets monthly. Two months on target means you can tighten. Three weeks of misses means finding the cause before touching the number, and the cause is usually peak staffing or unclear ownership. A launch, a new time zone or a jump in volume means running the staffing check again.
FAQ
What is a good first response time for live chat?
For a small team, a few minutes while chat is switched on is realistic, and 2 to 3 minutes is a reasonable first goal. Teams with dedicated chat staff aim for under a minute. Whatever you choose, write it as a share within a time and count only the hours chat is actually on.
Should automatic replies count as the first response?
Not in the number you manage by. They are worth sending because they set expectations, but they don't mean anyone has read the problem. Some tools count them by default, so check yours and keep a first human reply figure alongside it.
Should I aim for 80, 90 or 95 percent within target?
90 percent is a good default. At a small team, 95 percent usually means holding spare capacity all day, because one sick day or one long call pushes you under. Start at 90 and move the time before you move the percentage.
Do I need separate targets for evenings and weekends?
Only if someone works them. If nobody does, measure in business hours and tell customers when you will be back. If you do cover weekends, treat that as its own window with a named person on call and a looser target, rather than stretching the weekday numbers over it.
Should I tell customers what the target is?
Yes, in plain words and where they will see it: the contact page, the email auto-reply and the chat widget's offline message. "We reply within 4 business hours" gives a customer a time to plan around, which is easier to live with than an open-ended wait.
